encouragement, rent-free housing, and the help they gave me when it came to ...... in recorded battery voltages were also investigated and the old battery pack was ... with 2â wide white EZ Tear Construction Tape (Micronova Mfg., Inc., Part No. ...... recondition and recover their performance before dependable sensor pH ...
AN EVALUATION OF THE PERFORMANCE OF AN ISFET pH SENSOR IN A DYNAMIC ESTUARINE SYSTEM
by Stephen Fisher Gonski
A thesis submitted to the Faculty of the University of Delaware in partial fulfillment of the requirements for the degree of Master of Science in Marine Studies
Fall 2016
© 2016 Stephen Fisher Gonski All Rights Reserved
AN EVALUATION OF THE PERFORMANCE OF AN ISFET pH SENSOR IN A DYNAMIC ESTUARINE SYSTEM
by Stephen Fisher Gonski
Approved:
__________________________________________________________ Wei-Jun Cai, Ph.D. Professor in charge of thesis on behalf of the Advisory Committee
Approved:
__________________________________________________________ Mark A. Moline, Ph.D. Director of the School of Marine Science and Policy
Approved:
__________________________________________________________ Mohsen Badiey, Ph.D. Acting Dean of the College of Earth, Ocean, and Environment
Approved:
__________________________________________________________ Ann L. Ardis, Ph.D. Senior Vice Provost for Graduate and Professional Education
ACKNOWLEDGEMENTS
I would like to express my sincere gratitude to my advisor Dr. Wei-Jun Cai for his guidance and assistance throughout the entirety of my project. I am extremely grateful for everything he has taught me and the opportunities he has given me. I would also like to express my sincere gratitude to Dr. Karl Booksh and Dr. William Ullman for serving on my committee. Both have supported me throughout both my undergraduate and graduate studies at the University of Delaware. Dr. Booksh’s technical insights into the operational theory of chemical sensors was invaluable to my project. Dr. Ullman’s extensive working knowledge of the Murderkill Estuary-Delaware Bay system was indispensable, and his infectious enthusiasm for Oceanography and scientific discovery has passed on to me over the entire time I have known him. His mentorship has truly shaped my career. I would to like to thank Dr. Wei-Jen Huang, Dr. Chris Main, Dr. Tye Pettay, Dr. Najid Hussain, Dr. Baoshan Chen, and Dr. Janet Reimer for all their assistance during the various phases of my project. I would like to thank Dr. Todd Martz, Dr. Philip Bresnahan Jr., Dr. Yui Takeshita, and Taylor Wirth of the Scripps Institute of Oceanography for all their help with the various technical aspects of my project. I would to thank my Aunt Martha & Uncle Mark O’Neill for all the encouragement, rent-free housing, and the help they gave me when it came to building things for the duration of my Master’s work at the University of Delaware.
iii
I am extremely grateful to my lab mates for all their help throughout the course of my project, and I would like to thank my friends and family for all their support. Finally, I would like to thank Dr. Chris Winn and Dr. Sam Khang who first introduced me to seawater pH and set me on this course in 2013 when I worked with them at Hawai’i Pacific University prior to transferring back to UD.
iv
TABLE OF CONTENTS LIST OF TABLES ...................................................................................................... viii LIST OF FIGURES ....................................................................................................... ix ABSTRACT ................................................................................................................ xiii Chapter 1
INTRODUCTION .............................................................................................. 1
2
AN EVALUATION OF THE PERFORMANCE OF AN ISFET pH SENSOR IN A DYNAMIC ESTUARINE SYSTEM...................................... 10 2.1 2.2
Introduction ............................................................................................ 10 Materials & Methods .............................................................................. 16 2.2.1 2.2.2 2.2.3 2.2.4 2.2.5 2.2.6 2.2.7
Study Site................................................................................... 16 Deployment Design ................................................................... 17 Sensor Preparation ..................................................................... 19 Field Deployment ...................................................................... 20 Sampling Approach ................................................................... 22 Sensor Maintenance................................................................... 24 pH Calculation ........................................................................... 25 2.2.7.1 2.2.7.2
2.2.8
ISFET Operational Theory ........................................ 25 FET|INT and FET|EXT ............................................. 27
Calibration ................................................................................. 29 2.2.8.1 2.2.8.2 2.2.8.3 2.2.8.4 2.2.8.5
Field Measurements................................................... 31 Analytical Methods ................................................... 32 Independent Reference pH ........................................ 34 Departure from Open-Ocan Calibration Approach ... 35 Type of Calibrations .................................................. 36 2.2.8.5.1 2.2.8.5.2
2.2.9 2.3
pH Domian .............................................. 37 Raw Signal Domain ................................. 38
Assumptions & Limitations ....................................................... 41
Results .................................................................................................... 43 2.3.1
Discrete Sample pH Comparisons ............................................. 43
v
2.3.2 2.3.3
Raw Sensor pH vs. Independent Reference pH ......................... 46 * Calculating New Series of Esensor,f (T = 25℃) ....................... 47 2.3.3.1 2.3.3.2
2.3.4
Calibrated Sensor pH vs. Independent Reference pH ............... 50 2.3.4.1 2.3.4.2
2.4
Preliminary Comparisons .......................................... 48 * Setting Final Values of Esensor,f (T = 25℃) ............ 49
01 June 2016 .............................................................. 50 02 August 2016.......................................................... 52
Discussion............................................................................................... 53 2.4.1
Electrode Performance in an Estuarine System......................... 53 2.4.1.1 2.4.1.2 2.4.1.3 2.4.1.4
2.4.2
Effects of Excess Alkalinity ...................................................... 58 2.4.2.1 2.4.2.2 2.4.2.3 2.4.2.4
2.4.3
Excess Alkalinity Calculations .................................. 58 Excess Alkalinity in the Murderkill EstuaryDelaware Bay System................................................ 59 Quantifying Effects of Excess Alkalinity on Sensor Output ............................................................ 61 Recommendations for the Treatment of Excess Alkalinity in Future Work ......................................... 62
Electrode Conditioning in an Estuarine System ........................ 63 2.4.3.1 2.4.3.2
2.4.4
General Electrode Response ...................................... 53 Temporal Evolution of ∆pH INT-EXT Anomaly .......... 54 Quality Control Considerations ................................. 57 Sensor Drift ............................................................... 57
Eletrode Conditioning at Beginning of Sensor Deployment ............................................................... 63 Intra-Depployment Electrode Conditioning .............. 65
Recommendations ..................................................................... 67 2.4.4.1
Choice of Independent Referene pH ......................... 67 2.4.4.1.1
Murderill Estuary-Delaware Bay Results .................................................... 67
vi
2.4.4.1.2 2.4.4.2 2.4.5 3
Practical Considerations ......................... 68
Modifications to SeapHOx Design ............................ 72
Sensor Redundancy .................................................................... 75
SUMMARY & CONCLUSIONS .................................................................... 77
TABLES ....................................................................................................................... 79 FIGURES ..................................................................................................................... 86 REFERENCES ........................................................................................................... 125
vii
LIST OF TABLES Table 1
Deployment Details ................................................................................. 79
Table 2
Independent reference pH, salinity constraints, and alphanumeric designation for each sensor calibration method for each reference electrode for each sampling day .............................................................. 80
Table 3
∗ Average values of final calibration constants (Esensor,f ) applied to all raw sensor time series.............................................................................. 81
Table 4
Root-mean squared error (RMSE), sensor offset (c0), and sensor gain (c1) calculated from Model II least squares fits for all final sensor pH and reference pH comparisons from the 01 June 2016 sampling day ........................................................................................................... 82
Table 5
Root-mean squared error (RMSE), sensor offset (c0), and sensor gain (c1) calculated from Model II least squares fits for all final sensor pH and reference pH comparisons from the 02 August 2016 sampling day ........................................................................................................... 83
Table 6
Root-mean squared error (RMSE), sensor offset (c0), and sensor gain ′ ′ (c1) calculated from Model II least squares fits of pH sensor vs. pH disc property-property comparisons from the 01 June 2016 sampling day .... 84
Table 7
Root-mean squared error (RMSE), sensor offset (c0), and sensor gain ′ ′ (c1) calculated from Model II least squares fits of pH sensor vs. pH disc property-property comparisons from the 02 August 2016 sampling day ........................................................................................................... 85
viii
LIST OF FIGURES Figure 1
Different SeapHOx configurations used over the course of the project: Left Panel – v1.0, Middle Panel – v2.0, and Right Panel – v3.0 ............ 86
Figure 2
Example (Top Left Panel) and extent (Top Right Panel) of sediment accumulation inside SeapHOx flow housing experienced over 2 weeks during the 02 April 2016 to 26 August 2016 SeapHOx deployment. Unsuccessful preventive measures of outfitting flow housing with multiple outflow points (Bottom Left Panel) and employing a hydrodynamic filter stuffed with fiberglass mesh (Bottom Right Panel) to combat sediment accumulation inside SeapHOx flow housing .................................................................................................... 87
Figure 3
INT EXT Measured (A) pHraw and pHraw during periods of sediment accumulation inside SeapHOx flow housing and (B) turbidity from 21 April 2015 to 05 May 2015 ..................................................................... 88
Figure 4
Overview of SeapHOx field deployment method for September 2015 to August 2016 SeapHOx deployments. ................................................. 89
Figure 5
Examples of biofouling observed within a 1-2 week period experienced during SeapHOx deployments from summer 2016 (Left Panel) and summer 2015 (Right Panel) ................................................... 90
Figure 6
Functional implementation of the ISFET operating principle (from Martz et al., 2010) ................................................................................... 91
Figure 7
disc disc INT EXT pHraw , pHraw , calculated pHDIC−TA , and calculated pHelec on the total hydrogen ion concentration scale (pHT) as a function of salinity from 0900-1900 on the 01 June 2016 sampling day ............................... 92
Figure 8
disc disc INT EXT pHraw , pHraw , calculated pHDIC−TA , and calculated pHelec on the total hydrogen ion concentration scale (pHT) as a function of salinity from 0800-1930 on the 02 August 2016 sampling day ........................... 93
Figure 9
disc disc INT Property-property plot of pHraw vs. pHDIC−TA and pHelec as a function of salinity from the 01 June 2016 sampling day shown relative to a 1:1 (pH INT = pH disc ) relationship ..................................... 94
Figure 10
disc disc EXT Property-property plot of pHraw vs. pHDIC−TA and pHelec as a function of salinity from the 01 June 2016 sampling day shown relative to a 1:1 (pH EXT = pH disc ) relationship ..................................... 95
ix
Figure 11
disc disc INT Property-property plot of pHraw vs. pHDIC−TA and pHelec as a function of salinity from the 02 August 2016 sampling day shown relative to a 1:1 (pH INT = pH disc ) relationship ..................................... 96
Figure 12
disc disc EXT Property-property plot of pHraw vs. pHDIC−TA and pHelec as a function of salinity from the 02 August 2016 sampling day shown relative to a 1:1 (pH EXT = pH disc ) relationship ..................................... 97
Figure 13
disc ∗ Calculated values of EINT,1 (T = 25℃) from pHDIC−TA and disc ∗ EINT,2 (T = 25℃) from pHelec for all measurements as a function of salinity from the 01 June 2016 sampling day .......................................... 98
Figure 14
disc ∗ Calculated values of EEXT,1 (T = 25℃) from pHDIC−TA and disc ∗ EEXT,2 (T = 25℃) from pHelec for all measurements as a function of salinity from the 01 June 2016 sampling day .......................................... 99
Figure 15
disc ∗ Calculated values of EINT,1 (T = 25℃) from pHDIC−TA and disc ∗ EINT,2 (T = 25℃) from pHelec for all measurements as a function of salinity from the 02 Aug 2016 sampling day ........................................ 100
Figure 16
disc ∗ Calculated values of EEXT,1 (T = 25℃) from pHDIC−TA and disc ∗ EEXT,2 (T = 25℃) from pHelec for all measurements as a function of salinity from the 02 August 2016 sampling day.................................... 101
Figure 17
INT disc Property-property plot of pHfinal,1 vs. pHDIC−TA as a function of salinity from the 01 June 2016 sampling day ........................................ 102
Figure 18
EXT disc Property-property plot of pHfinal,1A vs. pHDIC−TA as a function of salinity from the 01 June 2016 sampling day ........................................ 103
Figure 19
EXT disc Property-property plot of pHfinal,1B vs. pHDIC−TA as a function of salinity from the 01 June 2016 sampling day ........................................ 104
Figure 20
INT disc Property-property plot of pHfinal,2 vs. pHelec as a function of salinity from the 01 June 2016 sampling day..................................................... 105
Figure 21
EXT disc Property-property plot of pHfinal,2A vs. pHelec as a function of salinity from the 01 June 2016 sampling day..................................................... 106
Figure 22
EXT disc Property-property plot of pHfinal,2B vs. pHelec as a function of salinity from the 01 June 2016 sampling day..................................................... 107
x
Figure 23
INT disc Property-property plot of pHfinal,1 vs. pHDIC−TA as a function of salinity from the 02 August 2016 sampling day.................................... 108
Figure 24
EXT disc Property-property plot of pHfinal,1 vs. pHDIC−TA as a function of salinity from the 02 August 2016 sampling day.................................... 109
Figure 25
INT disc Property-property plot of pHfinal,2 vs. pHelec as a function of salinity from the 02 August 2016 sampling day ................................................ 110
Figure 26
EXT disc Property-property plot of pHfinal,2 vs. pHelec as a function of salinity from the 02 August 2016 sampling day ................................................ 111
Figure 27
Murderkill Estuary-Delaware Bay System pH time-series from 09 May 2016 to 09 June 2016 .................................................................... 112
Figure 28
Murderkill Estuary-Delaware Bay System pH time-series from 20 July 2016 to 24 August 2016 ................................................................. 113
Figure 29
Calculated ∆pHINT−EXT anomaly shown relative to a zero ∆pH INT−EXT anomaly and salinity from 09 May 2016 to 09 June 2016 ....................................................................................................... 114
Figure 30
Calculated ∆pHINT−EXT anomaly shown relative to a zero ∆pH INT−EXT anomaly and salinity from 20 July 2016 to 24 August 2016 ....................................................................................................... 115
Figure 31
Contributions of excess alkalinity to total measured alkalinity on the 01 June 2016 sampling day ................................................................... 116
Figure 32
Contributions of excess alkalinity to total measured alkalinity on the 02 August 2016 sampling day ............................................................... 117
Figure 33
Composite of excess alkalinity effects time-series for 09 May 2016 to 09 June 2016 SeapHOx deployment ..................................................... 118
Figure 34
Composite of excess alkalinity effects time-series for 20 July 2016 to 24 August 2016 SeapHOx deployment ................................................. 119
Figure 35
Conditioning period at the start of the 09 May 2016 to 24 August 2016 SeapHOx deployment............................................................................ 120
Figure 36
Composite of intra-deployment conditioning time-series following sensor maintenance on 20 July 2016 ..................................................... 121
xi
Figure 37
Composite of intra-deployment conditioning time-series following sensor maintenance on 12 August 2016 ................................................ 122
Figure 38
INT disc Property-property plot of pHfinal,1 vs. pHDIC−TA as a function of salinity from the 01 June 2016 sampling day ........................................ 123
Figure 39
EXT disc Property-property plot of pHfinal,2 vs. pHelec as a function of salinity from the 02 August 2016 sampling day.................................... 124
′
′
xii
′
′
ABSTRACT
The feasibility of the Honeywell Durafet to the measurement of pH in a dynamic, productive, high-fouling, highly-turbid estuarine environment was investigated. Three different configurations of the SeapHOx sensor equipped with a Honeywell Durafet and its integrated internal (Ag/AgCl reference electrode containing a 4.5M KCl gel liquid junction) and external (solid-state chloride ion selective electrode, Cl-ISE) reference electrodes were deployed between April 2015 to August 2015 and September 2015 to August 2016 in the Murderkill Estuary-Delaware Bay System (Delaware, USA). Comprehensive sensor maintenance was performed every 1-2 weeks to ensure proper continuous sensor operation. Discrete bottle samples for dissolved inorganic carbon (DIC), total alkalinity (TA), and pHNBS were collected every 30 minutes to calibrate raw sensor output on more than 20 tidal cycle sampling trips. During these sampling trips, the full range of biogeochemical exchange between the fresher Murderkill Estuary outflow and the more saline Delaware Bay water endmembers was captured. The sensor pH collected during the summer of 2016 using the furthest refined SeapHOx configuration exhibited good agreement with the independent reference pH used. Accordingly, the sensor pH had a root-mean squared error (RMSE) ranging between 0.011 and 0.036 pH units across the full salinity range of the deployment environment relative to both pHT calculated from measured DIC and TA and pHNBS measured with a glass electrode corrected to pHT at in-situ conditions.
xiii
In this environment, the Honeywell Durafet proved capable of making highresolution, high-frequency pH measurements ranging between 6.8 and 8.4. Natural pH fluctuations ranging from 1 pH unit were routinely captured. A number of deficiencies in existing deployment guidelines and calibration protocol for Durafet-based biogeochemical sensors specific to estuarine and coastal ocean systems were identified and highlighted. Further, aspects of electrode response requiring further investigation were highlighted. A set of recommendations for the future utilization of these sensors in estuarine and coastal systems also resulted from the present work. The present work effectively demonstrated the viability of the Honeywell Durafet to the accurate and dependable measurement of pH as a part of future estuarine and coastal ocean CO2 chemistry studies. As such, this was a vital step in linking the parallel emerging trends in seawater pH metrology associated with the spectrophotometric measurement of pH using purified colorimetric indicator dyes and the electrochemical measurement of pH using the Honeywell Durafet over the full temperature and salinity range of natural waters. When these trends finally converge, the excellent accuracy of the Honeywell Durafet characterized under open-ocean conditions ( 20 where the standards are well calibrated. In addition, the convention of defining pH in terms of an activity or a concentration below this salinity threshold continues to be disputed. The effects of this discord propagate through all the major aspects associated with the measurement and use of pH in natural waters of S < 20. The standardization needed to govern and inform research incorporating the measurement and treatment of pH over this salinity range continues to elude the
8
scientific community (Dickson et al., 2016). Ultimately, a well-defined and traceable set of reference materials is needed to develop, calibrate, and test suitable pH measurement methodologies for natural waters of S < 20 (Dickson et al., 2016; Millero, 1986). It is only recently that a growing movement within the seawater pH metrology community seeking to develop a high-precision spectrophotometric pH measurement methodology using purified indicator dyes applicable over the full temperature and salinity range of natural waters has emerged. Spurred on by the findings of Yao et al. (2007), steps have been taken to incorporate purified indicator dyes in subsequently developed and refined spectrophotometric pH measurement methodologies (DeGrandpre et al., 2014; Lai et al., 2016; Liu et al., 2011; Patsavas et al., 2013a; 2013b; Soli et al., 2013). Concurrently, parallel work contributing to the development of such measurement methodologies that would facilitate pH measurements under near-zero temperatures (DeGrandpre et al., 2014; Papadimitriou et al., 2016), at much higher pressures (Hopkins et al., 2000; Rodriguez et al., 2015; Soli et al., 2013; Takeshita et al., 2016), and below open-ocean salinities (French et al., 2002; Gabriel et al., 2005; Hammer et al., 2014; Lai et al., 2016; Mosley et al., 2004; Yao and Byrne, 2001) has also been completed. The calibration of a new series of TRIS buffers prepared by varying the ratios of TRIS:TRIS-HCl added to them, using a Harned Cell, produced buffers of similar composition to natural waters of S < 20 that possess the required buffering capacity to perform the much needed method development for this task (Pratt, 2014).
9
Chapter 2 AN EVALUATION OF THE PERFORMANCE OF AN ISFET pH SENSOR IN A DYNAMIC ESTUARINE SYSTEM
2.1
Introduction In recent years, greater consideration has been given to the characterization of
pH in contemporary studies of the chemistry of natural waters. This type of work is central to investigations conducted over a full spectrum of temporal and spatial scales (Martz et al., 2010). An integral part of this work has encompassed the development of in-situ autonomous biogeochemical sensor packages capable of operating at high frequencies in remote areas with minimal user interaction over extended periods of time (weeks to months), over which continuous high-quality data of all types may be collected (Bresnahan et al., 2014; DeGrandpre et al., 1999; Martz et al., 2010). Measurements of this type help resolve the temporal and spatial patterns and trends that traditional discrete water sampling in many different environments is unable to capture (Ben-Yakov et al., 1974). Presently, there are two major classes of chemical sensors that are being utilized to accomplish this goal when it comes to pH: Reagentbased Optical Chemical Sensors (ROCS) and Electrochemical Sensors (ECS) (Hulanicki et al., 1991). ROCS employ an analyte-selective reagent (usually a pH-sensitive colorimetric indicator dye) to convert the analyte signal into an optical signal proportional to analyte concentration, whereas ECS transform the effect of the electrochemical interaction between the analyte-electrode couple into a useful signal
10
(Hulanicki et al., 1991). Each type of sensor has the potential to be adapted to autonomous measurements of pH. However, with the refinement of the spectrophotometric pH measurement methodologies using pH-sensitive colorimetric indicator dyes in the last 30 years (Byrne, 1987; Clayton and Byrne, 1993; Liu et al., 2011; Robert-Baldo et al., 1985; Zhang and Byrne, 1996) and its wide-spread use over electrochemical pH measurement methodologies, the development of these sensing systems has been steered in favor of ROCS over ECS (Martz et al., 2010). This trend has been amplified by the introduction of cost-effective and operationally stable optical components (such as waveguides and LED light sources) (DeGrandpre et al., 1999; Rérolle et al., 2013) and smaller, better-quality spectrophotometers capable of being integrated into various field-portable (Friis et al., 2004) and microfluidic (de Vargas Sansalvador et al., 2016; Rérolle et al., 2013) ROCS packages. The extensive literature of pH sensors based on ROCS designs developed in the last 10-15 years utilizing various optical properties like absorbance (de Vargas Sansalvador et al., 2016; Martz et al., 2003; Rérolle et al., 2013; Seidel et al., 2008) and fluorescence (Clarke et al., 2015) or luminescence (Larsen et al., 2011) using immobilized indicator dyes supports this view. Despite the prevailing trends exhibited in recent autonomous pH sensor development, ROCS incorporating pH-sensitive colorimetric indicator dyes are not without their limitations. Such sensors are composite systems that incorporate intricate arrangements of sensor components consisting of both bulkier hardware like pumps and valves and more delicate optical media like LEDs and wave guides that must
11
properly work in sequence over a prolonged measurement cycle to achieve the desired result. As such, some types of ROCS are known for their complicated operation and slow instrumental response times (Martz et al., 2010; Rérolle et al. 2016). Moreover, their elevated power requirements (Rérolle et al., 2016) and sometimes componentspecific operational temperature dependence (Bagshaw et al., 2016) makes them undesirable for long-term deployments (e.g. > 6 months) in some systems (Rérolle et al., 2016). Even more, the sensitivity of some ROCS to freeze-thaw cycles would limit their use in polar environments and at near-zero to sub-zero temperatures experienced in-situ in range of different deployment environments (Bagshaw et al., 2016). As an added caveat to ROCS operation, some types are subject to optical interferences (Liebsch et al., 2001) which could slow instrumental response times and compromise those optical measurements used to calculate pH. Ultimately, this could severely limit their use in highly turbid waters commonly found in estuarine and coastal marine ecosystems if the waters sampled are not filtered. Autonomous pH sensors based on ROCS designs are reputed to produce calibration-free measurements of pH given that they are only dependent on the optical properties inherent to the indicator dye used, thus resulting in repeatable, drift-free continuous measurements (DeGrandpre et al., 1999; Martz et al., 2003). However, since measurements consume indicator, this places a simple constraint on ROCS operation (Rérolle et al., 2016). Indicator dyes are also known to be less stable in solution form and subject to temporal degradation, as demonstrated in measurements of pH using their benchtop counterparts (Dickson et al., 2007). Indeed, this caveat of
12
ROCS operation would result in erroneous, unrepresentative measurements, which only further adds to their complexity and uncertain reliability. Additionally, ROCS operation is limited in the range of environmental temperature and salinity manifested in the constraints of the thermodynamic constants of the indicator dye (e.g. constants of purified mCP outlined in Liu et al. (2011)) used to calculate pH (Bagshaw et al., 2016; Rérolle et al., 2016). Likewise, indicator dyes used in ROCS pH measurements may have to be purified as well (Bagshaw et al., 2016; Rérolle et al., 2016) due to well-known pH errors attributed to dye impurities (e.g. mCP - Yao et al., 2007). Alternatively, microfluidic sensing systems offer significant operational advantages over conventional, bulky, power hungry ROCS packages (Bagshaw et al., 2016; Nightingale et al., 2015), and represent a promising future for pH sensors based on ROCS designs (Nightingale et al., 2015). Insofar, this technology has yet to achieve widespread use for the autonomous measurement of pH in natural waters of S < 20. A viable, robust, simple, and straightforward alternative to pH sensors based on ROCS designs was found in the form of a pH-sensitive Ion-Selective Field Effect Transistor (ISFET) (Martz et al., 2010). For a more detailed discussion of ISFET operational theory, see Section 2.2.7.1. The Honeywell Durafet is the current preeminent ISFET pH sensor used in open-ocean seawater CO2 chemistry and ocean acidification studies (Bresnahan et al., 2014; Martz et al., 2010). The Durafet is well adept to this purpose given its quick response time and low noise effects due to its low impedance (Martz et al., 2010), its stable and consistent linear response with temperature (Takeshita et al., 2014), and its
13
well-characterized signal stability at 1 atm (Sandifer and Voycheck, 1999). The Durafet is now paired with two quality reference electrodes, an Ag/AgCl reference electrode containing a 4.5M KCl gel liquid junction (internal) and a solid-state chloride ion selective electrode (Cl-ISE) (external) (Bresnahan et al., 2014; Martz et al., 2010). These characteristics justify the operation of the Honeywell Durafet on single calibration point over weeks to months without periodic recalibration. Operating under these circumstances, the Honeywell Durafet has repeatedly demonstrated a short-term precision (hours) and long-term precision (weeks to months) under controlled laboratory conditions of ±0.0005 pH units and ±0.005 pH units, respectively (Martz et al., 2010). Also, it routinely exhibited the desirable longterm precision of better than ±0.01 pH units under in-situ open-ocean conditions (Bresnahan et al., 2014). The published precisions of pH measurements made with the Honeywell Durafet at open-ocean salinities (S=30-36) are bracketed by the precision needed to perform general water quality work (±0.1 pH units) (Millero, 1986; Yao and Byrne, 2001), general estuarine chemistry work (±0.02 pH units) (Head, 1985), longterm ocean acidification monitoring (±0.001-0.003 pH units) (Bates, 2007; Byrne et al., 2010; Dore et al., 2009; González-Davila et al., 2007; Midorikawa et al., 2010; Newton et al., 2015) and to constrain the marine CO2 system in seawater (±0.001 pH units) (Yao and Byrne, 2001). When combined with the ease of use associated with its sensor packages and straightforward calibration approach (Bresnahan et al., 2014), the worldwide use of the Honeywell Durafet has grown very quickly (Frieder et al., 2012; Martz et al., 2014; Matson et al., 2011; Price et al., 2012).
14
The Honeywell Durafet and its dual-reference electrode configuration were originally integrated into two autonomous biogeochemical sensors known as the SeaFET and SeapHOx. The SeaFET is the passively flushed sensor that lacks an active pumping mechanism and is limited to the collection of pH data. The SeapHOx is the actively flushed sensor equipped with a Sea-Bird Electronics 5M submersible pump capable of making measurements of temperature, salinity (Sea-Bird Electronics Conductivity-Temperature Sensor – SBE37), pH, and oxygen (Aanderaa Data Instruments 4835 Optode). All of the SeapHOx components deploy sequentially in one flow stream (Bresnahan et al., 2014). Since the introduction of these sensor packages, the Honeywell Durafet has begun to be utilized for the measurement of pH over the full environmental range of temperature and salinity characteristic of natural waters using a variety of different platforms. Besides the use of the Honeywell Durafet under open-ocean conditions, it’s use for the measurement of pH at high pressures (Johnson et al., 2016), at near-zero temperatures (Bagshaw et al., 2016; Matson et al., 2011; Rérolle et al., 2016), and at low ionic strengths (Bagshaw in prep.; cited in Bagshaw et al., 2016) has been demonstrated or discussed. The Honeywell Durafet can also be used underway on hydrographic cruises (Rérolle et al., 2016), for profiling down to 3000 meters using the Deep Sea Durafet (Johnson et al., 2016), and on mobile oceanographic monitoring platforms such as the WavepHOx (Bresnahan et al., 2016) and Argo Floats (Johnson et al., 2016). The versatility of the Honeywell Durafet suggests that it will become the sensor of choice in the future to complement the application of a spectrophotometric
15
pH measurement methodology for dilute estuarine waters. Hence, there is a clear need to ascertain the feasibility of Durafet use in natural waters of S < 20 characteristic of estuarine and coastal ocean systems.
2.2
Materials & Methods
2.2.1
Study Site The Murderkill Estuary is a well-mixed, turbid, and nutrient-replete tributary
(Voynova et al., 2015) of the significantly larger Delaware Bay. The Delaware Bay is 215 km long, 18 km wide at its mouth up to about 44 km at widest point further upstream (Wong et al., 2009). It drains a 36,570 km2 watershed that includes large parts of Delaware, New Jersey, Pennsylvania, and the state of New York (Sharp et al., 2009). The Murderkill Estuary discharges into the Delaware Bay at Bowers, DE approximately 39 km upstream of the Bay mouth (Voynova et al., 2015). This small tributary estuary covers about 20 km between its discharge point at Bowers, DE and its upstream point at Frederica, DE. It has an average channel width of 50 meters and average depth of 4.5 meters (Wong et al., 2009). The two dominant system endmembers in the Murderkill Estuary-Delaware Bay System are the fresher Murderkill Estuary outflow and the more saline Delaware Bay water. The Murderkill River and Estuary and the Delaware Bay are both turbid due to elevated suspended sediment concentrations attributed to the numerous sources of suspended sediments in their watersheds, their tidal currents, and the erosion of ancient, muddy sedimentary deposits (deWitt & Daiber, 1974; Ullman et al., 2013; Voynova et al., 2015).
16
The Murderkill Estuary Watershed is composed of 275 km2, of which 55% is agricultural/farmland, 11% is covered by forests, 14% is urban, and 17% is covered by wetlands (comprised of 36% tidal wetlands and 60% freshwater wetlands) (DNREC, 2006). Moreover, the Kent County Regional Wastewater Treatment Facility (KCRWTF) which serves about 130,000 total customers drains into the Murderkill Estuary around 10.5 km upstream from its mouth. This subjects the estuary to greater population pressures than its predominantly agricultural watershed would suggest (KCBPW, 2012). It is important to note that the inputs from the watershed’s wetlands and marshes and inputs from the wastewater treatment plant comprise minor third and fourth endmembers in the Murderkill Estuary-Delaware Bay System (Dr. William Ullman, personal communication). Summed up, these sources contribute to high nutrient concentrations in this system that can fuel episodic phytoplankton blooms that can persist over periods of hours, days, or longer (Voynova et al., 2015), which provide ideal food sources for zooplankton and other estuarine heterotrophs (Cloern, 1996).
2.2.2
Deployment Design Over the course of the ten-month deployment period with SeapHOx SP053
from September 2015 to August 2016 and the preceding 4.5-month deployment with SeapHOx SP033, three different sensor configurations were used (Figure 1). Each new subsequent configuration incorporated modifications to sensor design meant to improve its applicability to the measurement of pH in an estuarine system. SeapHOx
17
Configuration v1.0 (SeapHOx SP033) (Figure 1, Left Panel) was simply the sensor as designed for open-ocean deployments following its assembly and testing at the Scripps Institute of Oceanography. This unit was deployed from 08 April 2015 to 26 August 2015 in the Murderkill Estuary-Delaware Bay System, but was removed in late August 2015 after sensor failure was detected. SeapHOx Configuration v2.0 (SeapHOx SP053) (Figure 1, Middle Panel) was equipped with a new flow path consisting of rounded tubing connections of a standardized cross-sectional area or inner diameter (R-3400 Tubing, 1/2” I.D. x 3/4” O.D., McMaster-Carr Part No. 5255K26) between the SB37, SBE 5M submersible pump, and flow housing that lacked all but one of the right-angle elbow connectors found on the original sensor flow path. The design of the SeapHOx SP053 flow path was further refined into SeapHOx Configuration v3.0 (Figure 1, Right Panel) by abandoning the remaining right-angle elbow connector at the outflow of the SBE37 conductivity cell and integrating a rounded tubing connection (R-3400 Tubing, 3/8” I.D. x 5/8” O.D, McMaster-Carr Part No. 5255K24) via a tight-seal barbed tube reduction fitting (1/2” I.D. x 3/4" O.D., McMaster-Carr Part No. 5463K639) to the 1/2” I.D. tubing on the remainder of the flow path. It must be noted that thicker tubing better retained its shape when rounded tubing connections were fashioned from it, which minimized constrictions in the new flow path. Because of the modifications, the flow path was considerably longer. Accordingly, a much longer flushing time was used for SeapHOx Configurations v2.0 and v3.0 (e.g. 60 seconds for 09 May 2016 to 09 June 2016 deployment and 70 seconds for 20 July 2016 and 24 August 2016 deployment) than what is recommended for open-ocean
18
deployments (25 seconds) (Martz, 2012). The increased flushing time substantially reduced the build-up of sediment and its associated fouling of the flow path of the instruments. Finally, a sampling interval of 30 minutes was chosen for the present work.
2.2.3
Sensor Preparation Prior to each sensor redeployment in the present work, sufficient steps were
taken to verify proper sensor operation and ensure the maintenance of such operation over periods of time between scheduled sensor maintenance trips. The necessary tasks conform with the Best Practices outlined in Bresnahan et al. (2014) and those detailed in a similar section of Rivest et al. (2016). Briefly, following sensor removal, the sensor was completely disassembled to facilitate comprehensive interior and exterior sensor cleaning procedures and the removal of biofouling organisms of all types. Trends in power consumption over the course of the previous deployment manifested in recorded battery voltages were also investigated and the old battery pack was replaced if warranted. Upon sensor reassembly, all O-rings were replaced (if needed) and regreased, all cable connectors were regreased, and any corroded or fouled screws found anywhere on the exterior of sensor were replaced. Following sensor reassembly, the SBE 5M submersible pump was removed, submerged, and tested to ensure proper operation. Optode operation was also tested in air pending the measurement of ~270 µM at ~100% oxygen saturation to verify performance against the original optode calibration as detailed in Bresnahan et al.
19
(2014). Using filtered seawater (salinity ~ 29-31), tests were conducted on the response of both electrodes to ensure measurements returned to within nominal voltage ranges for each reference electrode described in Bresnahan et al. (2014). Once all the tests of all sensor components were completed, the sensor body was wrapped with 2” wide white EZ Tear Construction Tape (Micronova Mfg., Inc., Part No. EZT2WH) followed by overlapping layers of copper tape. Copper tape was employed as the primary antifouling measure for the present work consistent with the recommendations of Bresnahan et al. (2014), while the construction tape greatly simplified the process of removing it after subjection to multiple months of in-situ estuarine conditions. The SeapHOx was also fitted with a replacement U-shaped copper pipe inlet at that time as well. Then, electrodes were conditioned in filtered seawater taken from the lower Delaware Bay while the sensor was continuously powered for the periods of time between sensor deployment periods. Electrodes remained stored in seawater during transport to the deployment site as well. For a more detailed evaluation of the applicability of the recommended open-ocean electrode conditioning protocol for a sensor deployments in an estuarine system, see Section 2.4.1.
2.2.4
Field Deployment For the duration of the sensor deployments, the SeapHOx units were deployed
alongside the suite of co-located sensors comprising the Kent County Land-Ocean Biogeochemical Observatory (LOBO) (http://kentcounty.loboviz.com/) (Table 1)
20
except during December 2015-March 2016 when those sensors were undergoing annual maintenance and conditioning. The LOBO deployment platform consisted of an aluminum cage with removable struts mounted on a trolley made of 1” thick galvanized steel raised and lowered by means of a marine winch onto an I-Beam situated on a pier adjacent to the mouth of the estuary that terminated with a concrete slab set 2.5 feet above the estuary floor. SeapHOx SP033 (SeapHOx Configuration v1.0) (Figure 1, Left Panel) was deployed upright. Thus, sensor performance suffered greatly from the effects of sediment accumulated over the electrode surfaces (Figure 2, Top Left/Top Right Panels) as indicated in the sustained pH decrease captured by both reference electrodes (Figure 3). Over the course of this deployment, various measures to combat sediment accumulation inside the flow housing were employed (Figure 2, Bottom Left/Bottom Right Panels) with only limited success. Ultimately, sediment was still able to accumulate over periods of variable turbidity conditions between sensor maintenance trips as it settled out of suspension from the waters left inside the flow housing after each sampling cycle once pumping ceased. When deploying subsequent SeapHOx configurations with SeapHOx SP053, the sensor was attached to the LOBO cage using a pair of aluminum brackets joined to the white SeapHOx brackets (Figure 4). A stainless-steel hose clamp (9-3/8” x 12-1/4” clamp ID, McMaster-Carr Part No. 5011T44) was also wrapped around the struts and attached sensor with pieces of gasket rubber separating the aluminum from the stainless steel as a fail-safe stabilization measure. Prior to deployment, the aluminum brackets and all aluminum LOBO components were coated with black Interlux
21
Pacifica Plus antifouling paint (Interlux Paint, LLC., Union, NJ, USA). To prevent a repeat of the SeapHOx SP033 failure, SeapHOx SP053 was deployed upside down so any sediment would accumulate away from the electrodes inside the flow housing. For future sensor deployments in turbid environments, an upside down or horizontal orientation that minimizes sediment effects is recommended. Likewise, either orientation is also more neutral with respect to the strain placed on the pump associated with pumping against gravity. The flow path was plumbed into the original outflow point of the flow housing to ensure it filled bottom-to-top in its new orientation keeping the electrodes and optode always immersed. In addition, the copper pipe inserts integrated as a secondary antifouling measure into the flow path of SeapHOx Configuration v1.0 were replaced by a U-shaped copper pipe inlet (Mueller Industries, MCTP-W Type ACR Refrigeration/AC Copper Pipe, 1/2" O.D.) bent to a 180o angle. The U-shaped inlet was placed at the sensor inflow point to prevent the passive settling of sediment and/or active invasion of fouling organisms into the sensor flow path. Because multiple metals were used for various parts of the deployment assembly, sufficient anti-corrosion measures were employed.
2.2.5
Sampling Approach For a sensor deployment in a dynamic, productive, high-fouling, highly-turbid
estuarine system characterized by high turbidity and large salinity changes over relatively short-time scales (Ullman et al., 2013; Voynova et al., 2015), the discrete sampling regiment adopted to calibrate the sensor should not be based on the static-
22
point approach practiced by operators of these sensors in open-ocean environments. Instead, the prevailing biogeochemical controls should be integrated into the discrete sampling regiment and discrete bottle samples should be collected over a period long enough to capture the full fluctuations in the parameters of interest in the deployment environment. If this dynamic-point approach is not practiced, sensor operators risk introducing unintentional bias into any subsequent sensor calibrations. The Murderkill Estuary-Delaware Bay System is essentially a river mouth and flood-dominant system (Dzwonkowski et al., 2013; Wong et al., 2009) characterized by significant discharge asymmetry resulting in much slower ebb tides and more rapid flood tides (Dzwonkowski et al., 2013). It is known that both high-frequency tidal fluctuations and low-frequency subtidal controls of spring-neap tides, winds, and large storm events drive local biogeochemistry (Dzwonkowski et al., 2013; Voynova et al., 2015; Wong et al., 2009). Thus, these biogeochemical controls were considered when planning sampling trips, which allowed for tidal cycle sampling to be adopted where discrete bottle samples were collected over the course of 10-12 hour periods. For this reason, the full range of salinity and pH associated with the periods of time when one or another system endmember dominated and periods of mixing between the two endmembers interspersed between were captured. Hence, this reduced the chances of introducing bias into reference pH datasets used to the perform the sensor calibration in terms of salinity, pH, or system endmember. Accordingly, sampling trips were also synchronized with favorable conditions related to subtidal controls of local biogeochemistry such as winds and the temporal responses of both system
23
endmembers to large storm events that represent significant freshwater inputs to the watershed (Voynova et al., 2015). This allowed the sensor operator to maximize the biogeochemical range the dynamic-point calibration covered. To this end, the incorporation of local biogeochemical controls into dynamic-point sensor calibrations performed on Durafet-based biogeochemical sensors must continue in the future in similar deployment environments when practical.
2.2.6
Sensor Maintenance Sensor maintenance was continuously performed every 7-10 days from May
2016 to August 2016, following previously compiled SOPs designed to minimize the effects of sediment and fouling organisms on sensor performance between maintenance trips. The degree of fouling observed on the SeapHOx over a 1-2 week period in an estuarine system (Figure 5) was comparable to that of much longer openocean SeapHOx deployments. Fouling appeared to be considerably worse from March to October coinciding with the most productive times of year in the Murderkill Estuary-Delaware Bay System (Ullman et al., 2013; Voynova et al., 2015), whereas it was only minimal over the winter months from November to February. SeapHOx maintenance was performed concurrently with that of the co-located Kent County LOBO equipment. This reduced the chances of degraded sensor performance attributed to fouling organisms or sediment not directly in contact with the SeapHOx, but could still indirectly affect data quality through altering the chemistry of the waters directly surrounding the instruments sampled by the SeapHOx.
24
2.2.7
pH Calculation The SeapHOx reports a pair of pH values based on the dual-reference electrode
design of this sensor package. These pH values and their associated calibration constants must be calculated from sensor voltages, in-situ temperature, and salinity measured over the course of each individual sampling cycle. In the subsequent section, a review of the ISFET operational principle and the derivations of the essential equations comprising the operation theory of each reference electrode is discussed.
2.2.7.1 ISFET Operational Theory Conventional ISFET sensors are a subdivision of Chemical Sensitized Field Effect Transistors (CHEMFETs) similar to a Metal Oxide Semiconductor Field Effect Transistor (MOSFET) technology (Hulanicki et al., 1991). However, an ISFET lacks the metal gate electrode over the conduction channel (Martz et al., 2010). ISFETs generally operate by applying a constant drain-source voltage and employing a feedback circuit to maintain it (Martz et al., 2010) (Figure 6). The metal gate electrode was replaced with a thin insulating layer of amphoteric material capable of facilitating both surficial protonation and deprotonation reactions that covers the surface of the conduction channel located between the source and the drain (Martz et al., 2010). The function of the metal gate electrode is satisfied through the direct exposure of the insulating material to the solution (Rérolle et al., 2012). The pH of the solution is then determined from voltage measurements between a reference electrode and the sensing layer (Martz et al., 2010) manifested in the change in the drain-source current usually
25
held constant (Hulanicki et al., 1991) as controlled by the site-binding protonationdeprotonation process occurring the solution-insulator interface (Martz et al., 2010). In mathematical terms, ISFET response can be calculated via the following equation (Janata, 2009):
∆VG =
2.303RT ai log ( + 1) z ⁄z zi F ∑j K i,j aj i j
(11)
where ΔVG is the ISFET response, R is the Gas Constant, T is the temperature in Kelvin, F is the Faraday Constant, zi is the charge of the ion of interest, zj is the charge of any interfering ions present, ai is the activity of the ion of interest, aj is the activity of any interfering ions present, and Ki,j are the intrinsic dissociation constants of the ion of interest and any interfering ions present, respectively. Like other ion-selective sensors, the measured change in the drain-source current or ISFET response (ΔVG) is proportional to the activity and concentration of the ion of interest in solution (Mikhelson, 2013) and hence, the pH of the solution (Janata, 2009). For a more comprehensive explanation on the possible effects of other interfering ions present in equation 11, see Section 2.2.9. For the application of the proton-sensitive Honeywell Durafet to the measurement of the pH of natural waters on a concentration scale using a conventional reference electrode that exhibits a Nernstian response to the free chloride ion, equation 11 can be rewritten into the general forms given below (Bresnahan et al., 2014): ∗ Esensor = Esensor −(
RT ln(10) × log(aH aCl )) F
26
(12)
RT ln(10) RT ln(10) ∗ Esensor = Esensor −( × log(𝛾H 𝛾Cl )) − ( × log(mH mCl )) (13) F F ∗ where Esensor is the measured voltage, Esensor is the calibration constant, ai is the
activity of either H + or Cl− , 𝛾i is the ion activity coefficient of either H + or Cl− , and mi is the molal concentration of either H + or Cl− . In addition, a nominal +1 gain added to the corresponding quantity is programmed into the operation of all sensors built around the Honeywell Durafet effectively bringing equations 12 and 13 in line with equation 11. Using the logarithm laws, equation 13 can be rearranged to isolate pH or the − log(mH ) as seen below: RT ln(10) ∗ Esensor = Esensor −( × [log(𝛾H 𝛾Cl mCl ) − pH]) F
(14)
Temperature-dependent standard potentials for the reference electrode configuration incorporating the Honeywell Durafet are instead referred to as calibration constants ∗ (Bresnahan et al., 2014) designated by an asterisk (*) (e.g. Esensor ) over a nought o symbol (o) (e.g. Esensor ) because of the inter-sensor variability observed in these
quantities characterized from previous work (Martz et al., 2010).
2.2.7.2 FET|INT and FET|EXT As designed, the SeapHOx can make two separate pH measurements using its two integrated reference electrodes – FET|INT (Ag/AgCl) and FET|EXT (Cl-ISE) – designated by the notations of pH INT and pH EXT , respectively (Martz et al., 2010). In
27
the case of FET|INT, the necessary parameters denoted by ‘INT’ can be calculated via the following equations (Martz et al., 2010): ∗ EINT = Ej + EINT − S × log(mH )
(15)
∗ EINT = {E ∗ (FET|INT) − S × log[𝛾H (sw)] − S × log[aCl (ref gel)]}
(16)
pH
INT
∗ ) (EINT − EINT = S
(17)
where S is the Nernst slope (S = {RTln(10)}/F); Ej is the liquid junction potential; aCl is the chloride ion activity of the KCl reference gel in FET|INT; and 𝛾H and mH represent the ion activity coefficient and molality of the hydrogen ion in solution. Since FET|INT can measure dissolved H+ in solution and dissolved Cl- in the reference gel, it always has a repeatable chloride ion activity (aCl ). Because of this, log(𝛾H 𝛾Cl mCl ) is incorporated into the operational definition of the voltage converted to pH via the temperature-dependent Nernst slope in equation 17 (Martz et al., 2010). In the case of FET|EXT, it does not demonstrate a repeatable aCl (Martz et al., 2010). To a similar effect, a set of equations detailed in Martz et al., (2010) can be used to calculate the key parameters as denoted by ‘EXT’ as well: EEXT = E ∗ (FET|INT) − S × log(𝛾H 𝛾Cl ) − S × log(mCl ) − S × log(mH ) ∗ EEXT = E ∗ (FET|EXT)
pH
EXT
∗ ) (EEXT − EEXT + S × log(𝛾H 𝛾Cl mCl ) = S
(18) (19) (20)
where S is the Nernst slope (S = {RTln(10)}/F) and 𝛾i and mi represent activity coefficients and molality of either H + or Cl− , respectively. Since FET|EXT measures
28
dissolved hydrogen chloride (HCl) in solution, the mean activity coefficient of hydrogen chloride (HCl) (𝛾± (HCl)2 = 𝛾H 𝛾Cl ) calculated from temperature and salinity (Khoo et al., 1977) and the concentration of the free chloride ion in seawater (mCl ) calculated from salinity (Dickson et al., 2007) are used to calculate pH EXT (Bresnahan et al., 2014; Martz et al., 2010). Using the measured voltage, in-situ temperature, and salinity, pH INT and pH EXT are calculated assuming a 100% Nernst ∗ slope (e.g. 59.16 mV/pH at 25oC) and a constant d𝐸𝑠𝑒𝑛𝑠𝑜𝑟 /d𝑇 (e.g. ∗ ⁄ ∗ ⁄ d𝐸𝐼𝑁𝑇 d𝑇 = −0.001101 V/℃ 𝑎𝑛𝑑 d𝐸𝐸𝑋𝑇 d𝑇 = −0.001048 V/℃) (Bresnahan et
al., 2014; Martz et al., 2010).
2.2.8
Calibration Conventional calibration protocol for electrochemical pH measurement
methods consist of employing a series of traceable reference materials (e.g. TRIS Buffers in ASW) representative of the composition of natural waters at those salinities to calibrate and periodically recalibrate the electrodes or sensors to be used. When performing an electrochemical pH calibration using one of the three concentration scales for work in low and intermediate salinities, it has been recommended that a single standard reference buffer at an intermediate salinity between freshwater and seawater (e.g. 15 < S < 20) or a series of standard buffers spanning the full salinity range encountered be used (Whitfield et al., 1985). Yet, the problem with the former is that the errors associated with any measurements made greatly increase when deviating from the single calibration salinity (Whitfield et al., 1985), while the
29
problem with the latter rests in the difficulties in the preparation and preservation of each individual buffer (Easley and Byrne, 2012). In the case of the latter, it has been demonstrated that a 10-unit change in the salinity between the seawater buffer solution and natural seawater sample yields errors of up to 0.028 pH units and even a difference of only 2 salinity units results in errors of 0.005 pH units (Easley and Byrne, 2012). Certainly, this highlights the need for finer salinity incrementation between successive buffers and only adds to the complexity of this method (Easley and Byrne, 2012; Whitfield et al. 1985). The difficulties of standard buffer calibration methods for electrochemical pH determinations below S=20 were circumvented by employing the calibration method for Durafet-based biogeochemical sensors recommended by Bresnahan et al. (2014). This type of sensor calibration is dependent on the natural variability inherent to the deployment environment (Hofmann et al., 2011; Kline et al., 2012) rather than a comparison against standard reference buffers. Such an in-situ or field calibration is performed through the collection of in-situ pH data via a series of discrete samples coinciding with the time of sensor measurements measured using well-established benchtop methods and then calibrating a working Durafet to those field measurements (Bresnahan et al., 2014). Using this approach, the primary control on the quality of the sensor calibration and subsequent sensor time series is directly related to the amount and quality of the discrete samples collected over the course of a sensor deployment. An optimal discrete sampling regiment is characterized by >10 usable discrete bottle
30
samples (Rivest et al., 2016). On the other hand, calibrating a sensor time series to only one discrete bottle sample can produce sensor inaccuracies of ~0.1 pH units (Bresnahan et al., 2016) and is strictly recommended against (Bresnahan et al., 2014) due to the limitations it places on the evaluation of sensor failure or effects of biofouling (Rivest et al., 2016). A second control is related to the benchtop pH measurement method chosen to measure the pH of the discrete bottle samples Past experiences dictate that the pH measurements of discrete bottle samples from openocean deployments be performed using spectrophotometric pH measurement methods using pH-sensitive colorimetric indicator dyes (Rivest et al., 2016) known for their high precision, simple analytical procedures, and minimal operator interaction (Carter et al., 2013; DeGrandpre et al., 2014). On the other hand, the application of spectrophotometric pH measurement methodologies is not straightforward in lower salinity waters as previously discussed in Section 2.1. Therefore, this limits the applicability of spectrophotometric pH measurement methods to sensor deployments in natural waters below S=20. Yet, other pH measurement methodologies exist and pH can be calculated from two marine CO2 system parameters – dissolved inorganic carbon (DIC), partial pressure of CO2 (pCO2), and total alkalinity (TA) – so a sensor calibration in an estuarine environment is still achievable.
2.2.8.1 Field Measurements Dissolved inorganic carbon (DIC), total alkalinity (TA), and pHNBS were measured at the sensor deployment site located at the confluence of the Murderkill
31
Estuary and Delaware Bay at Bowers Beach, DE on 01 June 2016 from 0900-1900 and 02 August 2016 from 0800-1930. A 1200-Watt gasoline-powered generator (Champion Power Equipment) was used to supply power to all equipment over the course of these sampling trips. Water column samples were continuously collected every 30 minutes coinciding with sensor measurements (00:00 and 00:30) made at the same frequency using a peristaltic pump by way of a 12.5’ length of tubing lowered into the water and sufficiently weighted to maintain a fixed position. The tubing was placed at the approximate location of the sensor relative to the pier its deployment structure was mounted on. Measurements of in-situ temperature, salinity, and pressure were taken from the SeapHOx.
2.2.8.2 Analytical Methods Samples for DIC and TA were collected usually in duplicate following filtration through Whatman 0.45 µm Polyethersulfone (PES) filters (GE Healthcare Bio-Sciences, Pittsburgh, PA, USA) by bottom-filling into triple-rinsed 250-mL borosilicate bottles. The samples were then fixed with 100 µL of saturated mercuric chloride (HgCl2) solution. Approximately every 30 minutes, filters were regenerated by rinsing the filter backwards then forwards three times using a reservoir of deionized (DI) water and a second peristaltic pump following individual sampling cycles. Upon returning from the field, the samples were preserved in 4oC for future analysis (Cai and Wang, 1998; Jiang et al., 2008). DIC was determined through acid extraction by quantifying the released CO2 using an infrared gas analyzer (AS-C3 Apollo Scitech).
32
TA was measured by Gran Titration (Gran, 1950; 1952) using a semi-automated opencell titration system (Cai et al., 2010; Huang et al., 2012). All measurements were calibrated against certified reference materials (CRM, provided by A.G. Dickson from Scripps Institute of Oceanography). Samples for pHNBS were collected unfiltered always in duplicate via bottomfilling into triple-rinsed 125-mL clear glass bottles. pHNBS was measured using an Orion Dual Star pH/ISE Benchtop Meter equipped with a new Orion 8302BNUMD Ross Ultra Glass Triode pH/ATC Combination Electrode (Thermo Fisher Scientific Inc., Beverly, MA, USA) within 3-5 minutes of sample collection at in-situ temperature. The pH electrode was calibrated every 2-3 hours using three National Bureau of Standards (NBS) traceable freshwater pH buffers of 4.01, 7.00, and 10.01 stored in 20 mL scintillation vials thermostatted in water taken from the Murderkill Estuary for at least 30 minutes prior to the calibration. During all pHNBS measurements, parafilm was used to minimize CO2 -exchange between the sample and the surrounding atmosphere. Also, all pH measurements were performed in the shade under an umbrella to minimize electrode drift caused by exposure to direct sunlight. Fresh 20 mL aliquots of the pH buffers were also used for each sampling trip to minimize the errors in pHNBS measurements potentially imparted by the pH buffers themselves.
33
2.2.8.3 Independent Reference pH From the measurements of DIC, TA, and pHNBS , two sets of independent reference pH emerged that were used to calibrate the raw sensor time series. The first independent reference pH was pHT calculated from measured DIC and TA at in-situ disc temperature, salinity, and pressure (pHDIC−TA ) using the inorganic carbon dissociation
constants from Millero et al. (2006), the bisulfate ion acidity constant of Dickson (1990), and the boron-to-chlorinity ratio of Lee et al., (2010) in the Excel macro CO2SYS (Pierrot et al., 2006). These constants were used for all other CO2SYS calculations performed. Calculations of pHT using the inorganic carbon dissociation constants from Mehrbach et al., 1973 refit by Dickson and Millero (1987) were also performed. However, preliminary results of the present work found that the constants from Millero et al. (2006) were more suited to calculations of pHT in estuarine waters (S < 30). The second independent reference pH was pHT calculated from the field measurements of pHNBS made on discrete bottle samples with a pH electrode. To use these pH measurements, they were first corrected to the SeapHOx temperature by applying the difference between values of pHNBS calculated from measured DIC and TA at the measurement temperature and at the SeapHOx temperature to them. This set field of pHNBS values was designated pHNBS . Following this step, a second correction
based on the difference between pHNBS and pHT calculated from measured DIC and TA at in-situ temperature (T), salinity (S), and pressure (P) solely attributed to the
34
field difference in the pH scales (∆pHscales ) was then applied to pHNBS to generate the disc second independent reference pH (pHelec ) per the following conventions: DIC−TA (T, S, P) − pHTDIC−TA (T, S, P) ∆pHscales (T, S, P) = pHNBS
(21)
disc field (T, S, P) = pHNBS (T, S, P) − ∆pHscales (T, S, P) pHelec
(22)
disc By applying this pair of corrections, the pHelec was converted to the total scale (pHT )
used in the sensor calibration and in-situ environmental conditions (temperature, salinity, and pressure) were conserved.
2.2.8.4 Departure from Open-Ocean Calibration Approach The remoteness of open-ocean sensor deployment sites can limit access to the sensor over the course of long-term deployments and the ability to collect quality discrete bottle samples. This often results in the need for complicated, time-intensive collection methods for those samples (e.g. via Niskin Bottle by Scuba Diver) (Rivest et al., 2016). It would also prevent the performance of regularly scheduled maintenance needed for the continuous collection of high-quality data in dynamic, productive, high-fouling, highly-turbid environments. Thus, tradition has dictated that a sensor time series generated from such deployments be treated as one continuous time series from beginning to end with pre-deployment calibrations and/or postdeployment calibrations sometimes attached to validate sensor performance (Bresnahan et al., 2014). In the case of an estuarine sensor deployment in an accessible location characterized by an adequate number of quality independent
35
reference pH measurements from quality discrete bottle samples and regularly scheduled sensor maintenance, a departure from this convention is justified. The results of such sensor deployments leave the sensor operator with many shorter-duration sensor time series that may be calibrated independently since the sensor operation is essentially reset following the conclusion of each maintenance session. These independent sensor time series can be combined after the sensor deployment to build an aggregated sensor time series. This more frequent and rigorous calibration approach provides a better way of accounting for the natural variability inherent to the estuarine deployment environment. Ultimately, this approach would facilitate an evaluation of the performance of the Honeywell Durafet in an estuarine environment over timescales bracketing those performed in controlled laboratory environments (hours to days) (Bresnahan et al., 2014; Martz et al., 2010) and in the open ocean (multiple months) (Bresnahan et al., 2014; Rivest et al., 2016).
2.2.8.5 Types of Calibrations At present, there are two contrasting methods that have been used to calibrate Durafet-based biogeochemical sensors: pH Domain and Raw Signal Domain. The first calibration method is comparative and only requires access to the sensor pH (pH sensor) and the independent reference pH. The second method is based on the raw signal domain of the data which is much more intricate and requires intimate knowledge of the sensor operational theory. The raw signal domain requires access to
36
the measured voltages and the independent reference pH as well. In the following sections, we provide a brief overview of each calibration type.
2.2.8.5.1 pH Domain Essentially, the calibration method encompassing the pH domain is a simple comparison of the raw, uncalibrated sensor output (pH sensor) against an independent reference pH (e.g. pH measured from discrete bottle samples - pH disc ). Propertyproperty plots of pH sensor (dependent variable) vs. pH disc (independent variable) provide valuable insight into sensor performance reflected in the degree of deviation of the sensor output from the truest pH of the water seen by the sensor at that time (pH disc ) (Bresnahan et al., 2014). These deviations can be characterized from the intercept (sensor offset - 𝑐0 ) and slope (sensor gain - 𝑐1) (Bresnahan et al., 2016) generated from a Model II least squares fit of the data (Peltzer, 2007). If significant deviations in the intercept and slope from 0 and 1, respectively, are observed, this suggests bias in the sensor and/or reference pH used in the comparison. Calibrating a raw sensor time series only using this approach is not recommended as it limits the identification of and the differentiation between all the potential sources of error, and may introduce significant error into the final sensor time series (Bresnahan et al., 2014).
37
2.2.8.5.2 Raw Signal Domain The raw signal domain provides a more robust and dependable means of calibrating a raw sensor output. It is done through the calculation and application of a ∗ new set of calibration constants (Esensor ) to displace the raw sensor measurements to
bring them in-line with the chosen reference pH (Bresnahan et al., 2014). To report a reasonable range of pH values for a given deployment environment, all SeapHOx units are programmed with a pair of initial calibration constants corrected to the calibration temperature of 25oC specific to each reference electrode when they are first ∗ ∗ assembled, EINT,0 (T = 25℃) (Ag/AgCl) and EEXT,0 (T = 25℃) (Cl-ISE). These
values serve as nominal placeholders used to calculate pH INT and pH EXT using ∗ equations 15-17 and 18-20, respectively (Martz, 2012). The values of EINT,0 (T = ∗ 25℃) and EEXT,0 (T = 25℃) programmed into SeapHOx SP053 were -0.4347 V and
-1.4070 V, respectively. Many researchers elect to perform pre-deployment calibrations, thus yielding a more dependable and realistic set of calibration constants that can be utilized from the onset of sensor deployments (Bresnahan et al., 2014). However, since this was a pilot deployment of a Durafet-based biogeochemical sensor in an estuarine environment with no available comparisons, a system of rigorous insitu sensor deployment calibrations was pursued. The calculation of new sets of calibration constants is simplified via the use of a MATLAB script that uses inputs of the measured sensor voltages (EINT (T)/ EEXT (T)), in-situ temperature (T), salinity, and the independent reference pH at in-situ
38
temperature (pHT (T)/pHF (T)) corresponding to the same time stamp of an individual sensor measurement. For the FET|INT, the new calibration constants at in-situ ∗ (T), temperature, EINT and the final calibration constants corrected to 25oC, ∗ (T = 25℃), are calculated via the following equations (Bresnahan et al., 2014): EINT,f ∗ (T) EINT = EINT (T) − S(T) × pHT (T) ∗ ∗ (T) (T = 25℃) = EINT EINT,f +
∗ d𝐸𝐼𝑁𝑇 (25℃ − T) d𝑇
(23) (24)
∗ ⁄ where S(T) corresponds to the Nernst slope at in-situ temperature and d𝐸𝐼𝑁𝑇 d𝑇 ∗ (T) corresponds to its previously defined quantity. For the FET|EXT, EEXT and ∗ (T = 25℃) are calculated as follows (Bresnahan et al., 2014): EEXT,f ∗ (T) EEXT = EEXT (T) + S × log(𝛾H 𝛾Cl mCl ) − S × pHF (T) ∗ ∗ (T) (T = 25℃) = EEXT EEXT,f +
∗ d𝐸𝐸𝑋𝑇 (25℃ − T) d𝑇
(25) (26)
where 𝛾i represents the ion activity coefficients of either H + or Cl− , mi represents the ∗ ⁄ molality of either H + or Cl− , and d𝐸𝐸𝑋𝑇 d𝑇 corresponds to its previously defined ∗ ∗ quantity. Throughout this process, values of EEXT (T) and EEXT,f (T = 25℃) are
calculated from independent reference values of pHF corresponding to the time stamp of an individual sensor measurement (Bresnahan et al., 2014). ∗ ∗ (T = 25℃) and EEXT,f (T = 25℃) are set to average values based Once EINT,f
on all valid discrete samples, the application of these values and subsequent recalculation of all pH values is expedited by a second MATLAB script designed to batch process all sensor measurements in a single sensor time series (Bresnahan et al.,
39
2014). The pH values are recalculated using modified forms of equations 17 and 20 for pH INT and pH EXT , respectively as follows: INT pHfinal
EXT pHfinal
=
=
∗ (T = 25℃)) (EINT − EINT,f
S
∗ (T = 25℃)) + S × log(𝛾H 𝛾Cl mCl ) (EEXT − EEXT,f
S
(27)
(28)
INT EXT where pHfinal and pHfinal correspond to the final calibrated pH values measured with
the internal and external reference electrodes, respectively. In this final step, standard practice dictates that the pHF taken from EEXT be converted to pHT, thereby making the two pH values measured by the different reference electrodes directly comparable (Bresnahan et al., 2014). When pH values populating a final sensor time series are calibrated, all raw sensor measurements (voltages, in-situ temperature, and salinity) are constants and used to recalculate the calibration constants based on the reference pH chosen. This means the calibration constants and the pH values are dependent one another such that the difference between the raw sensor pH and the independent reference pH is the principle control on the magnitude of the new calibration constant. This caveat propagates through the calibration process since the only thing that changes when calibrating a sensor time series is the calibration constant. So, the difference between sensor sensor the final calibrated sensor pH (pHfinal ) and raw uncalibrated sensor pH (pHraw ) is
dependent on and inversely related to the difference between the initial ∗ ∗ (T = 25℃)) and final (Esensor,f (T = 25℃)) calibration constants (Esensor,0
40
∗ (T = 25℃)) in equations 27 and 28. according to the quantity (Esensor − Esensor,f
When redefining the calibration constant and calibrating a sensor time series, the only change is the contribution of the calibration constant to the total voltage, ∗ (T = 25℃)), converted to pH via the temperature-dependent (Esensor − Esensor,f
Nernst slope. Since both calibration constants are negative quantities, if ∗ ∗ ∗ (T = 25℃) < Esensor,0 (T = 25℃), then (Esensor − Esensor,f (T = Esensor,f sensor ∗ sensor (T = 25℃)) and pHfinal 25℃)) > (Esensor − Esensor,0 > pHraw , and vice versa.
2.2.9
Assumptions & Limitations The deployment of Durafet-based biogeochemical sensors in an estuarine
environment experiencing lower salinities than those in open-ocean environments (S > 30) is not without its assumptions. The main assumption associated with the present work was assuming 100% Nernstian response in pH calculations over the full range of salinity encountered. Previous work has demonstrated that the dual-reference electrode configuration is capable of repeatedly exhibiting 100% Nernstian response over a wide range of pH (2-12) and a narrow salinity range (20-35) in seawater media (Takeshita et al., 2014). It was also assumed the effects of the major interfering cations found in seawater referred to in equation 11 are negligible (leading to pH error of 20 data.
2.3.4.2 02 August 2016 INT EXT The calibrated sensor pH of pHfinal,1 (Figure 23), pHfinal,1 (Figure 24), INT EXT pHfinal,2 (Figure 25), and pHfinal,2 (Figure 26) from 02 August 2016 also generally
exhibited good agreement when compared against their respective set of independent reference pH with sensor offsets (𝑐0 ) close to 0 and sensor gain (𝑐1 ) close to 1 (Table 5). Unlike the previously discussed sampling day, 02 August 2016 was characterized higher salinities (S > 20), a narrower salinity range, and slower rates of salinity disc change. Raw sensor time series recalibrated using pHDIC−TA exhibited greater sensor disc gains than those recalibrated with pHelec . This may be indicative of the effects of
excess alkalinity like what was observed on 01 June 2016. However, when the sensor INT EXT offsets and gains of pHfinal were compared with pHfinal , they were much closer to 0
and 1, respectively. This may indicate bias present in the sensor time series of the internal reference electrode attributed to the previously discussed sources.
52
2.4
Discussion
2.4.1
Electrode Performance in an Estuarine System A single pair of calibrated sensor time series from each sampling day
INT EXT calibrated using the same independent reference pH - pHfinal,2 + pHfinal,2A for 01 June INT EXT 2016 and pHfinal,2 + pHfinal,2 for 02 August 2016 – was used in lieu of the six total disc pairs. The sensor time series calibrated using pHelec were used because the time series
were always characterized by better Model II fit parameters and were theoretically free of the effects of excess alkalinity.
2.4.1.1 General Electrode Performance For the duration of the sensor deployments from 9 May 2016 to 09 June 2016 calibrated using the discrete bottle samples collected on 01 June 2016 (Figure 27A) and 20 July 2016 to 24 August 2016 calibrated using the discrete bottle samples on 02 August 2016 (Figure 28A), voltages measured with both the internal and external reference electrodes align very well across the full spectrum of salinity characteristic of a flood-dominant, estuarine system. Likewise, the pH values calculated from the measured voltages exhibit good agreement across the lower and more intermediate salinities encountered in the 09 May 2016 to 09 June 2016 deployment (Figure 27B) of S=3.25-25.00 and the consistently higher salinities characterizing the 20 July 2016 to 24 August 2016 deployment (Figure 28B) of S=15.33-29.83. During these two sensor deployments, pH fluctuations ranging from 1 pH unit were
53
routinely captured over the course of single tidal cycles over that full salinity range. Notable departures in the agreement between both the voltages and the subsequently calculated pH values were observed during periods of pH change during the ebb and flood tides. These departures were most noticeable at the minima of the profiles where both pH and salinity were generally at their lowest values, and at the maxima of the profiles when both pH and salinity returned to their higher values on the flood tide. The nominal seawater voltage ranges for EINT and EEXT characteristic of openocean conditions were reported to be 0.03 to 0.1 V and -0.95 to -0.80 V, respectively (Bresnahan et al., 2014). When compared, the voltage ranges for EINT characterizing the two estuarine deployments were much lower at around -0.025 to 0.055 V. This was to be expected in an estuarine system characterized by consistently lower salinities and sub-pH 7 values at low tide which resulted in lower measured voltages for EINT. On the other hand, when the voltage ranges of EEXT were compared, the minima fell slightly outside of the nominal range for seawater at -0.98 to -0.86 V. Similarly, a different voltage range for EEXT would be expected for an estuarine system that experiences a lower salinity range and more substantial simultaneous temperature, salinity, and pH variability.
2.4.1.2 Temporal Evolution of ∆𝐩𝐇 𝐈𝐍𝐓−𝐄𝐗𝐓 Anomaly The dual-reference electrode configuration incorporated into Durafet-based biogeochemical sensors like the SeapHOx is not an absolute necessity for their use in different environments (Bresnahan et al., 2014). In open-ocean deployment
54
environments, this configuration provides a simple, yet powerful method of detecting the effects of fouling or sensor failure through the temporal evolution of ∆pH INT−EXT anomalies (Bresnahan et al., 2014, Rivest et al., 2016). In an estuarine system, both the instantaneous salinity experienced and the relative change in salinity between successive sensor measurements can negatively impact electrode. The ∆pH INT−EXT anomaly is the only quasi-suitable metric used to gauge sensor performance over time to look for the effects of drift, fouling, and sensor failure in the present work as well. Average non-zero ∆pH INT−EXT anomalies were observed during both the 09 May 2016 to 09 June 2016 (0.006±0.063 pH units) and 20 July 2016 to 24 August 2016 (-0.008±0.020 pH units) deployments with conditioning periods removed. The ∆pH INT−EXT anomalies observed in the present work were much greater than those characterized by Bresnahan et al. (2014) ( pHfinal )
was consistently observed under prolonged periods of salinity decrease on the ebb tide which bottomed out at the lowest salinities coinciding with the greatest influences of the fresher Murderkill Estuary outflow. A rapid, sharp decrease to a negative INT EXT ∆pH INT−EXT anomaly (pHfinal < pHfinal ) coinciding with the tide change and the first
55
flood tide measurements followed. Afterwards, the ∆pH INT−EXT anomaly approached a near-zero value as the higher salinity Delaware Bay water inundated the deployment environment for extended periods of time prior to the next ebb tide. The result is an asymptotic profile for each tidal excursion across the full range of salinity experienced during both sensor deployments. The magnitude of the ∆pH INT−EXT anomalies was directly related to the magnitude of the salinity fluctuation, but these quantities were inversely related with respect to the exhibited trends. As characterized by Bresnahan et al. (2014) under open-ocean salinities, the source of this anomaly is most like a salinity lag arising from the inadequate flushing of the instrument flow path and flow housing. In other words, the salinity of the water sampled by the rapidly flushed conductivitytemperature sensor (SBE37) was different than the salinity of the water seen by the electrodes inside the more slowly flushed flow housing (Bresnahan et al., 2014). Also, this anomaly may also be attributed to a slower response time of one or both reference electrodes at S < 20 and/or under large, rapid salinity changes. This would be manifested in the inadequate reconditioning or re-equilibration of the reference electrodes to the characteristic salinities experienced under dynamic flow conditions in an estuarine system. This demonstrates the need to characterize the performance of the dual-reference electrode configuration at the extremes of environmental conditions typical of estuarine and coastal ocean systems in a controlled laboratory setting.
56
2.4.1.3 Quality Control Considerations Operators of Durafet-based biogeochemical sensors routinely report their Quality Control (QC) procedures used to flag, and if necessary, exclude data from final datasets incorporated into publications (Bresnahan et al., 2014; Matson et al., 2011; Rivest et al., 2016) and major data repositories (Rivest et al., 2016). These procedures use the measured pH ranges (Rivest et al., 2016) and variability of pH with time (Bresnahan et al., 2014; Rivest et al., 2016) as data quality tools. Measurements found to fall outside of accepted pH ranges of the deployment environment may be flagged and excluded. If the sensor pH is changing significantly faster than in-situ pH determined from other methods, sensor measurements may also be flagged and excluded as well (Rivest et al., 2016). For sensor deployments carried out in wellstudied open-ocean systems, these QC procedures are reasonable (Bresnahan et al., 2014; Rivest et al., 2016). In rapidly changing estuarine systems, these QC procedures are less justifiable. Due to the physical and chemical variability in our estuarine system, we have chosen to report all sensor measurements to inform future work done in similar environments. However, the performance of future sensor deployments in similar systems under similar conditions would help develop a set of recommended QC procedures moving forward.
2.4.1.4 Sensor Drift In a dynamic, productive, high-fouling, highly-turbid estuarine system characterized by a broad range of in-situ environmental conditions, there is a greater
57
possibility for sustained sensor drift that could be attributed to any number of sources (Bresnahan et al., 2014; Martz et al., 2010). However, due to the limitations of the experimental design, it would be difficult to attribute any anomalies observed in the data solely to sensor drift based on the complexity of sensor operation in this type of setting. To definitively attribute anomalies witnessed in the data to sensor drift, measurements or calculations of additional sets of independent reference pH made over a comparable frequency and range of environmental conditions would be needed. The current sensor deployment suffered from the fact that the only means of pH comparison available for the sensor output was by means of comparison against pH measured or calculated from discrete bottle samples collected over a significantly lower frequency. Even so, the accessibility of the deployment site allowed for the performance of regularly scheduled comprehensive sensor maintenance that should limit the occurrence, longevity, and magnitude of sensor drift in the present work. This was evident from the large pH fluctuations that were consistently captured throughout the present sensor deployment.
2.4.2
Effects of Excess Alkalinity
2.4.2.1 Excess Alkalinity Calculations In estuarine and coastal ocean systems, the effects of excess alkalinity, which is alkalinity present in an aquatic system not accounted for by the marine CO2 system has been well-characterized (Cai et al., 1998; Yang et al., 2015). Contributions to excess alkalinity can come in the form of organic material derived from humic
58
particles (Cai et al., 1998; Yang et al., 2015) as well as planktonic and bacterial cells (Kim & Lee, 2009; Ko et al., 2016). Excess alkalinity is often referred to as organic alkalinity if it proves to strongly correlate with total dissolved organic matter (DOM), total dissolved inorganic carbon (DOC), or a similar metric (Cai et al., 1998; Kim & Lee, 2009). Excess alkalinity is manifested as the difference in alkalinity (∆TA) between measured alkalinity (TAmeas ) and alkalinity calculated from two other field measured marine CO2 system parameters, which for the present work are pHNBS and pH−DIC
DIC, (TAcalc
): pH−DIC
∆TA = TAmeas − TAcalc
(29)
When excess alkalinity is present, any inorganic parameter calculated using the measured alkalinity, such as pHT calculated from measured DIC and TA will yield overestimations of them (Cai et al., 1998; Patsavas et al., 2015). This can be especially problematic when performing an intercomparison of marine carbonate chemistry measurements for a specific aquatic system (Patsavas et al., 2015). However, even when excess alkalinity is present, it may not always have a statistically significant effect on any subsequent marine CO2 system analyses (Loucaides et al., 2012; RibasRibas et al., 2014).
2.4.2.2 Excess Alkalinity in the Murderkill Estuary-Delaware Bay System The apparent effect of excess alkalinity on the two sets of independent disc reference pH was clearly distinguishable with values of pHDIC−TA consistently greater
59
disc than pHelec for all usable calibration points on both 01 June 2016 and 02 August 2016.
When calculated, average values of ∆TA on 01 June 2016 (Figure 31A) and 02 August 2016 (Figure 32A) were 12.6±1.8 µmol kg-1 and 14.7±2.8 µmol kg-1, respectively. Contributions of excess alkalinity generally increased with decreasing salinity which disc disc led to greater overestimations of pHDIC−TA relative to pHelec (Figures 31B and 32B)
when measured total alkalinities were lower. This aligned well with trends described from comparisons of sensor pH and independent reference pH in preceding sections. However, excess alkalinity was not consistently strongly correlated with salinity over the course of sampling on 01 June 2016 (r2 = 0.6564) or 02 August 2016 (r2 = 0.2390), nor with DOC calculated from fluorescence using the summer 2013 relationship from Voynova et al. (2015) for the same deployment site on 01 June 2016 (r2 = 0.0787) or 02 August 2016 (r2 = 0.0832). Nevertheless, these results aligned well with those of other marine CO2 chemistry studies performed in other parts of the Murderkill Estuary watershed near its confluence with the Delaware Bay (Ullman & Aufdenkampe, unpublished data). Excess alkalinity was also calculated using a constant mole-ratio of excess alkalinity to total dissolved organic carbon (DOC) (Patsavas et al., 2015). To do this, a constant organic-base-to-DOC ratio of 0.122 mol-alk/mol-C (Cai et al., 1998) and DOC calculated from fluorescence (Voynova et al., 2015) were used. Excess alkalinity in this system did not correlate well with the excess alkalinity calculated from DOC using this ratio on 01 June 2016 (r2 = 0.0787) or 02 August 2016 (r2 = 0.0832). No such mole ratio of excess alkalinity to DOC would suffice to describe the
60
contributions of excess alkalinity for this system. When summed up, it is possible that excess alkalinity is not related to total DOC but to a specific, uncharacterized fraction of the total and/or there is more than one source of excess alkalinity present in this system. Excess alkalinity never exceeded more than 1-2% of total measured alkalinity. The effects of excess alkalinity on the relative trends in pH observed in this system are insignificant when pH fluctuations as large as 0.5 pH units to a full pH unit are routinely experienced over tidal excursions alone.
2.4.2.3 Quantifying Effects of Excess Alkalinity on Sensor Output ∗ (T = 25℃) to an average value to minimize the When setting Esensor,f disc anomaly between the sensor pH and pHDIC−TA , the effects of excess alkalinity
captured in discrete bottle samples are averaged out. However, this approach does disc produce multiple calibrated sensor time series with those calibrated using pHDIC−TA disc exhibiting elevated pH signals relative to those calibrated using pHelec . Given the sensor ∗ (T = 25℃) and pHfinal inverse, dependent relationship between Esensor,f , the
magnitude of the difference between the corresponding pH values in each sensor time disc disc series calibrated using pHDIC−TA and pHelec will be controlled by the difference in the ∗ (T = 25℃) used for calibration. values of Esensor,f
Moreover, the temperature response of the Honeywell Durafet is partly controlled by the dependence of the calibration constant with temperature (directly ∗ (T = 25℃)) (Martz et al., 2010). Because of this, the small degree related to Esensor,f
61
sensor and nature of the variability characterizing the differences between values of pHfinal disc disc calibrated using pHDIC−TA and pHelec over time will depend on the relationship ∗ (T = 25℃) and temperature. These relationships are governed by between Esensor,f INT EXT ∗ ⁄ ∗ ⁄ d𝐸𝐼𝑁𝑇 d𝑇 and d𝐸𝐸𝑋𝑇 d𝑇 in equations 24 and 26 for pHfinal and pHfinal , respectively.
For the 01 June 2016 calibration (Figure 33), the difference between the values of ∗ INT (T = 25℃) was 0.0016 V, which translated into a difference in pHfinal EINT,f of
0.0268-0.0281 pH units. For the 02 August 2016 calibration (Figure 34), the ∗ (T = 25℃) was 0.0026 V, which translated difference between the values of EINT,f INT into a difference in pHfinal of 0.0429-0.0439 pH units. In both cases, the difference INT between the values of pHfinal matched up against a profile indicative of an inverse ∗ relationship with temperature. This adequately describes relationship between EINT EXT and temperature characterized by Martz et al. (2010). Similar comparisons of pHfinal disc disc calibrated using pHDIC−TA and pHelec produced similar results since the equations ∗ EXT (T = 25℃) and pHfinal used for EEXT,f are of an equivalent type and the same
calibration protocols were employed.
2.4.2.4 Recommendations for the Treatment of Excess Alkalinity in Future Work Our results show that existing calibration protocols for Durafet-based biogeochemical sensors cannot be used to detect, quantify, and correct for contributions of excess alkalinity to sensor measured pH values in estuarine and coastal ocean systems. Sampling for DOC and/or DOM and measuring additional
62
marine CO2 system parameters (e.g. DIC or pCO2) may be needed to gauge the magnitude of the impacts excess alkalinity has on pH measurements in these settings. If pH calculated from measured alkalinity and a second marine CO2 system parameter serves as the primary reference pH, it is recommended that all values be corrected for the effects of excess alkalinity prior to performing the sensor calibration. If a strong correlation exists between excess alkalinity and another measured parameter such as DOC, DOM, or salinity measured with comparable frequency, such a correction should be feasible.
2.4.3
Electrode Conditioning in an Estuarine System
2.4.3.1 Electrode Conditioning at Beginning of Sensor Deployment Conditioning periods at the beginning of sensor deployments can endure for a few hours to several days depending on the pre-deployment procedures undertaken to address electrode conditioning. If left unconditioned prior to deployment, the external reference electrode takes significantly longer to condition than the internal reference electrode due to the time needed for Br − ions to replace Cl− ions in the AgCl solid solution found in the Cl-ISE (Bresnahan et al., 2014). To minimize these conditioning periods, electrodes should be stored in seawater, be continuously powered for 5-10 days prior to deployment, and should remain stored in seawater while the instrument is transported to the deployment site as well (Bresnahan et al., 2014). These recommended procedures were used leading up to the redeployment of the sensor on 09 May 2016. Filtered seawater taken from the lower Delaware Bay with
63
a salinity range of S=29-31 throughout the year was used as the conditioning medium. The seawater was stored inside the flow housing of the SeapHOx from 13 April 2016 to 09 May and was replaced every 5 days. Without an independent reference pH measured or calculated with a comparable frequency, the ∆pH INT−EXT anomaly was used to gauge the success of the electrode conditioning procedure. A conditioning period of 74 hours followed (Figure 35) at which time, the calculated ∆pH INT−EXT anomaly returned to a consistently stable value of < ±0.01 pH units while the more saline Delaware Bay water inundated the deployment site for extended periods of time. The causes of this prolonged conditioning period were mostly likely attributed to the time needed to accomplish a stable flow of ions across the liquid junction of the internal reference electrode and the previous discussed replacement of Cl− ions with Br − ions in the solid AgCl solution of the external reference electrode as outlined in Bresnahan et al. (2014). With salinities ranging from 5.9 < S < 23.1, the implications of large changes in [Br − ] and [Cl− ] accompanying such large salinity fluctuations could result in a conditioning period of this length in an estuarine system. If the initial ion replacement process inside the Cl-ISE was slower than the tidal action, it is certainly possible that the conditioning period required for pH INT was much shorter since the conditioning signal was dominated by pH EXT . For this reason, the development of a suitable pre-deployment electrode conditioning protocol for sensor deployments in estuarine and coastal ocean systems is needed.
64
2.4.3.2 Intra-Deployment Electrode Conditioning At any time when the electrodes are not immersed and allowed to dry out, a conditioning period of varying length will always be required for the electrodes to recondition and recover their performance before dependable sensor pH measurements are again made (Bresnahan et al., 2014). The implications of these shorter, more numerous conditioning periods on data loss can be severe when regularly scheduled sensor maintenance was performed when the sensor was out of the water for 1-3 hours. Throughout the course of present work, the sensor was redeployed with – (1) only air in the flow housing or (2) filtered seawater stored in the flow housing – to determine the time required for the electrodes to recondition after each sensor maintenance trip. This was done by comparing sensor pH and pHT calculated from Cond measured DIC and TA (pHDIC−TA ) of discrete bottle samples collected following
sensor maintenance trips. The results from 20 July 2016 (Figure 36) following the sensor redeployment with filtered seawater stored in the flow housing and from 12 August 2016 (Figure 37) with only air stored in the flow housing were both promising. On 20 July 2016, discrete bottle samples were collected up to 6.5 hours after the first sensor measurement at 1200 at the end of a flood tide until the start of the following flood tide. In contrast, on 12 August 2016, discrete bottle samples were collected up to 6 hours after the first sensor measurement at 1130 (end of the ebb tide) until the point in time where the deployment site was inundated with the higher salinity Delaware Bay water. In both cases, a near-zero ∆pH Sensor−Cond anomaly was achieved for both
65
reference electrodes. Because of this, it can reasonably be assumed that under either condition, the reference electrodes adequately recondition within 6 hours of the first sampling cycle after sensor redeployment following sensor maintenance trips in the summer. Accordingly, a nominal 6-hour reconditioning period was assumed for all May 2016 to August 2016 data used in subsequent analyses for the present work. Given that near-zero ∆pH Sensor−Cond anomalies were achieved within 3.5 hours of the first sensor measurement after redeployment under both conditions, the shorter conditioning periods most likely arise from the fact the sensors simply needed to be reimmersed and made “wet”. Incidentally, the departure from zero ∆pH Sensor−Cond anomalies demonstrated the difficulties of constraining intradeployment conditioning periods with respect to each reference electrode in a dynamic estuarine environment. These difficulties relate to the challenge of collecting quality discrete bottle samples over changing tidal conditions and the presence of the salinity lag addressed in Section 2.4.1.2. These intra-deployment conditioning periods are most likely heavily dependent on local temperature and salinity conditions at the deployment site at different times of the year. Hence, this would make their length site-dependent and deployment-specific. It is recommended that filtered seawater or natural waters taken from the deployment site be stored in the flow housing of a SeapHOx before its redeployment following sensor maintenance. This is recommended because a narrower range for the ∆pH Sensor−Cond anomalies were yielded under this condition and it conforms with existing recommended electrode conditioning protocol (Bresnahan et al., 2014).
66
2.4.4
Recommendations
2.4.4.1 Choice of Independent Reference pH 2.4.4.1.1 Murderkill Estuary-Delaware Bay Results sensor The property-property plots of pHfinal vs. pH disc emphasize strong disc agreement between the sensor pH and reference pH. Our results show that pHelec
provided a suitable means of calibrating raw sensor output. This is supported by the sensor gains and offsets describing their comparisons yielding deviations less than or disc equal to their counterparts calibrated using pHDIC−TA from 1 and 0, respectively. This
indicates that comparable or reduced bias was present in comparisons made with disc disc disc pHelec versus pHDIC−TA . Calibrating sensor output against pHelec worked to a EXT INT disproportionately better effect on pHfinal than pHfinal . This is attributed to the fact
that the ∆pHscales decreases with decreasing salinity, so in effect, both pH EXT and ∆pHscales possessed some degree of salinity sensitivity. Elevated values of the calculated sensor offsets yielded from Model II least sensor squares fits of pHfinal vs. pH disc often make them more abstract and difficult to
directly compare. Instead, property-property plots between the perturbations in pH disc sensor and pHfinal from the minimum value of pH disc characterized by perturbation ′
′
sensor sensor variables of pH disc = pH disc − min(pH disc ) and pHfinal = pHfinal −
min(pH disc ) scale Model II fit parameters to the pH range of a dataset making them more representative (Bresnahan et al., 2016). Such property-property plots (Figures 38 and 39) produce Model II fit sensor offsets much closer to 0 (Tables 6 and 7).
67
Consistent with Bresnahan et al. (2016), comparisons of the results of the present work were made against the results of evaluations of seven widely-used pH sensors by the Alliance for Coastal Technologies (ACT, 2012: http://www.actus.info/evaluations.php) completed using pH perturbation property-property plots. The ACT results produced standard deviations in pH sensor − pH disc anomalies of 0.01-0.1 pH units under in-situ conditions. In the case of the present work, good agreement sensor between all pairs of pHfinal and pH disc values were demonstrated and similar sensor
validation protocol yielded standard deviations or root mean squared errors (RMSE) of 0.01-0.04 pH units across both reference electrodes over a range of environmental conditions. This demonstrates the viability of the Honeywell Durafet to the collection of dependable high-frequency, high-resolution, high-precision pH data over periods of weeks to months in dynamic, productive, high-fouling, highly-turbid estuarine and coastal ocean systems regardless of the independent reference pH used.
2.4.4.1.2 Practical Considerations disc disc Our results indicate that both pHDIC−TA and pHelec provided an adequate
means of calibrating raw sensor output. Both were used to help confirm aspects of electrode response at higher salinities (S > 20) characterized from previous work (Bresnahan et al., 2014; Bresnahan et al., 2016; Martz et al., 2010; Takeshita et al., 2014) and to describe electrode performance in an estuarine system. Moreover, our INT results of pHfinal also demonstrate the Honeywell Durafet continues to exhibit near-
68
ideal Nernstian response to at least S=8.99 well below the S=20 threshold characterized by Takeshita et al. (2014). In addition, the elevated particle loads characterizing the Murderkill Estuary-Delaware Bay System (Ullman et al., 2013) do not appear to exert a significant and measurable influence on instantaneous sensor sensor disc measurements given the similar agreement of pHfinal with pHDIC−TA (filtered) and disc pHelec (unfiltered). This may also need to be studied further. It is difficult to
determine long-term effects of turbidity on electrode response without comparisons against an independent reference pH measured over a similar frequency. In this deployment environment, it is known that the total suspended solids concentrations tend to be higher under stronger tidal currents, over prolonged periods of strong winds, and following strong localized storm events (Voynova et al., 2015). So, any long-term effects of turbidity on electrode performance may also be subject to similar controls. However, turbidity appears to not significantly affect the relative trends in pH captured over the course of the present work. The calibration of raw sensor output to pH calculated from measured DIC and TA at in-situ conditions has proven to work in the past (Bresnahan et al., 2016; disc Rérolle et al., 2016). However, the potential overestimation of pHDIC−TA due to
contributions of excess alkalinity in estuarine and coastal ocean deployment environments cannot be ignored. Sets of independent reference pH generated from pH calculated from two other measured marine CO2 system parameters can be used to calibrate a Honeywell Durafet in the future. This approach also offers the advantage of
69
providing the sensor operators with a better understanding of the underlying controls on the marine CO2 system in a specific environment (Macleod et al., 2015) via its constraint and its variability over a range of different environmental conditions (Bresnahan et al., 2016). Employing the calculation of pHT from two measured marine CO2 system parameters as a means of calibration also allows the sensor operators to maximize the number of parameter combinations used to calculate an independent reference pH, to better constrain electrode response, and to improve the quality of the calibrations. While this is beneficial, it is not always as practical given the labor-intensive requirement that such a calibration presents (Bresnahan et al., 2016). In the case of the present sensor deployment, a total of 486 discrete bottle samples for DIC and TA were collected on more than 20 sampling days throughout the year. These efforts produced 255 calibration points over the ten-month sensor deployment from September 2015 to August 2016. This translated into over 125 person hours attributed to sampling alone and does not include time dedicated to other concomitant tasks. This sort of comprehensive sensor calibration method is not always possible. In contrast, since marine CO2 system parameters can be measured more accurately than they can be calculated, calibrating a working Honeywell Durafet to direct measurements of pH in an estuarine system is a better approach. For the present work, an independent reference pH traceable to direct measurements of pHNBS made with a glass electrode calibrated with freshwater pH buffers was satisfactorily used to calibrate a Honeywell Durafet. While there will be uncertainty in any pH measurement
70
associated with using a standard with a fixed salt concentration as a reference for all pH measurements, under the pretext that the pHNBS measurements are free of electrode drift and biofouling, their correction to pHT at in-situ conditions substantially reduces this uncertainty. Given the well-documented problems associated with using freshwater pH buffers in seawater pH applications, this calibration method has been largely overlooked. More importantly, the low cost and wide availability of freshwater pH buffers provide a cost-effective and straightforward, yet robust calibration method for future sensor deployments in estuarine and coastal ocean systems. The calibration method chosen for a sensor deployment in an estuarine or coastal ocean system is subject an array of practical considerations. These considerations can be heavily deployment-specific and governed by the deployment site chosen (e.g. accessibility), resources available to the sensor operators (e.g. funding or instrumentation), available manpower, and time that can be spent on the different phases of the sensor deployment. To assist with the planning and execution of the present sensor deployments, the set of published Best Practices for the use of Durafetbased biogeochemical sensors outlined in Bresnahan et al. (2014) was indispensable, and the workflow plan for pH sensor deployments detailed in Rivest et al. (2016) should further streamline them in the future. Since the independent reference pH used in the present work was collected over a much lower frequency, it is imperative that future sensor deployments in estuarine and coastal ocean systems include the deployment of co-located,
71
independent sensors used to measure an additional marine CO2 system parameter (e.g. pCO2) and/or utilize regional empirical marine CO2 system relationships (e.g. Alin et al. (2012) for the southern California Current System) as recommended by Bresnahan et al. (2014). This would generate the set of high-frequency independent reference pH that is needed to examine the long-term trends in electrode response in this new type of environment. The present work clearly indicates that the employment of multiple sets of independent reference pH are a necessary tool for distinguishing between bias in sensor pH and reference pH in-line with the recommendations of Bresnahan et al. (2014). The choice of calibration method does not have not have a universal answer, but all aspects of future sensor deployments must carefully be considered before deciding on the type and breadth of calibration method to be used and the design of the discrete sampling regiment with which to carry it out.
2.4.4.2 Modifications to SeapHOx Design As discussed previously in Section 2.2.2, the open-ocean design of the SeapHOx was not practical in a dynamic, productive, high-fouling, highly-turbid estuarine system. As designed, the SeapHOx was not capable of dependable highresolution, high-frequency pH measurements over periods of days to weeks between sensor maintenance trips to validate its application to such an environment that was needed for the present work. A second SeapHOx unit modified for use in this type of deployment environment was successfully deployed for a period of about 10 months between September 2015 to August 2016. The results of the present work yielded a
72
comprehensive set of recommendations to be considered for a subsequent sensor design specific to estuarine and coastal ocean deployment environments. The effects of high turbidity have the potential to introduce bias not attributed to degraded electrode response into raw sensor output and subsequently calibrated sensor time series. This may lead to the flagging and exclusion of large numbers of sensor measurements after undergoing routine QC procedures. The magnitude of data loss attributed to high turbidity increases significantly in the absence of regularly scheduled sensor maintenance. In the present work, the degradation of pump performance over time was experienced due to the formation of choke points in the instrument flow path especially around areas where the cross-sectional area or the direction of the flow path changed. These features lead to progressively shorter periods of slower, weaker water flow through the instrument. By employing the flow path modifications discussed in Section 2.2.4, the periods of time characterized by full pump operation and strong water flow through the instrument is significantly extended. So, employing similar modifications on SeapHOx flow paths is recommended for their future use in turbid waters. Even with a modified flow path, particles were still routinely observed inside the impeller chamber of SBE 5M submersible pump. The presence of these particles around the plane of impeller rotation can be especially problematic. Given the variable pump performance experienced in this sensor deployment, it is recommended that SeapHOx units designed for use in estuaries and the coastal ocean be equipped with a stronger pump more adept to pumping waters with high particle loads. Ultimately, the Sea-Bird
73
Electronics (SBE) 5P submersible pump (Sea-Bird Scientific, Bellevue, WA) should prove well suited to this purpose if the stronger, faster flow through the flow housing of the SeapHOx does not compromise of the integrity of the sensing surfaces. Currently, the communication with any SeapHOx unit is limited by way of a direct RS-232 connection with a computer through a suitable terminal program such as HyperTerminal or Tera Term. On the other hand, the recently developed and tested mobile oceanographic sensing platform incorporating the dual-reference electrode design integrated into the SeapHOx known as the WavepHOx described by Bresnahan et al. (2016) achieved methods of wireless communication between the instrument and the sensor operator using an iPhone/iPad. Moving forward, it is recommended that subsequently designed SeapHOx units for use in deployment environments in close enough proximity to land (typical most potential estuarine and coastal ocean environments) within range of telecommunication networks (e.g., 4G/3G networks) and/or WiFi networks adept at facilitating remote communication between sensor and sensor operator incorporate wireless communication capabilities as well. With respect to the present work, wireless communications to help stop and start deployment loops, to cycle the sensor in and out of sleep mode, and to download sensor data remotely would have been preferred over removing the sensor from the water each time and manually performing all these tasks. The ability to download data remotely would also minimize data loss attributed to intra-deployment conditioning that occurred each time after sensor removal. Wireless communications would allow the sensor operator increased capability to monitor aggregate sensor performance throughout a
74
deployment, thus greatly improving the chances of success in challenging deployment environments like the one used in the present work. The current open-ocean design of the SeapHOx is characterized by the absence of technology that allows data to be sent and viewed by the sensor operations, PIs, and stakeholders remotely in real time. There exist data loggers and communications systems to improve the capability to monitor underwater instruments. The STOR-X Submersible Data Logger (Satlantic L.P., Halifax, Nova Scotia, Canada), integrated into the co-located suite of biogeochemical sensors of the Kent County Land-Ocean Biogeochemical Observatory (http://kentcounty.loboviz.com/), is capable of and has proven extremely useful in monitoring and maintaining underwater instruments during long-term deployments in the same deployment environment (Dr. William Ullman, personal communication). The benefits of action with respect to these tasks greatly outweighs the costs of inaction as the focuses of ocean acidification and marine CO2 chemistry research further expand into the less well-characterized dynamic estuarine and coastal ocean systems (Andersson et al., 2015; Cai, 2011; Duarte et al., 2013; Weisberg et al., 2016).
2.4.5
Sensor Redundancy Related work with the Honeywell Durafet over comparable timescales carried
out under rigorously-controlled laboratory conditions established small degrees of characteristic inter-sensor variability in key parameters associated with sensor ∗ ⁄ ∗ ⁄ operation (e.g. d𝐸𝐼𝑁𝑇 d𝑇 , d𝐸𝐸𝑋𝑇 d𝑇 ) (Martz et al., 2010). Because of this, sensor
75
redundancy (the use of multiple units) was a key feature in subsequent work (Bresnahan et al., 2014; Takeshita et al., 2014). For the present work, one SeapHOx unit with one Honeywell Durafet was used in a highly variable estuarine system, and the success of this sensor was verified using a pair of validation procedures. Further work is needed to properly constrain and interpret electrode response under the conditions experienced over the present sensor deployment to sufficiently substantiate its results.
76
Chapter 3 SUMMARY & CONCLUSIONS
Pilot deployments of the SeapHOx sensor equipped with the Honeywell Durafet were carried out between April 2015 to August 2015 and September 2015 to August 2016 in the Murderkill Estuary-Delaware Bay System (Delaware, USA). The present work encompassed the operation of a Honeywell Durafet in a dynamic, productive, high-fouling, highly-turbid estuarine system. In this setting, salinities lower than the characteristic open-ocean ranges (S=30-36) under which sensor performance had previously been investigated (Bresnahan et al., 2014) ranging between S=3.25 and S=29.33 were consistently experienced. Sensor performance was also evaluated over the dynamic salinity conditions on tidal excursions that characterize this system. The sensor pH collected during May 2016 to August 2016 using the furthest refined SeapHOx configuration and discrete sampling strategy exhibited good agreement with the independent reference pH used. The sensor pH was characterized by a root-mean squared error (RMSE) ranging between 0.011 and 0.036 pH units across the full environmental salinity range relative to both pHT calculated from measured DIC and TA and pHNBS corrected to pHT at in-situ conditions. Despite the challenges of operating a SeapHOx in an estuarine system, the results were very promising. In this environment, the Honeywell Durafet was capable of consistently capturing natural pH fluctuations nominally ranging from 1 pH unit
77
occurring in the surrounding waters across the more saline Delaware Bay water and the fresher Murderkill Estuary outflow endmembers mixed along the salinity gradient over a range of timescales. Sensor performance demonstrated the versatility of the Honeywell Durafet and reinforced its viability to the measurement of pH as a part of future estuarine and coastal ocean CO2 chemistry studies. A number of deficiencies in existing deployment guidelines and calibration protocol for Durafet-based biogeochemical sensors were identified and highlighted in this work. Aspects of electrode response requiring further investigation were also highlighted. Moreover, a comprehensive set of recommendations for the future utilization of these sensors in similar systems resulted from the present work. As the emerging parallel trends in seawater pH metrology for natural waters of S < 20 converge, the highly desirable accuracy of the Honeywell Durafet achieved under open-ocean conditions of better than 0.01 pH units should be attainable in estuarine and coastal ocean systems in the near-future as well.
78
TABLES
Table 1. Deployment Details.
79
Site Murderkill EstuaryDelaware Bay System (Bowers Beach, DE)
Regime FloodDominant Estuarine
Lat. 39.05 o N
Lon. 75.39 o W
Deployment Dates 08 April 2015 to 26 Aug 2015
Sensor Packages SeapHOx SP033, LOBO Sensors
Configuration v1.0
Validation Approaches Not Applicable due to Sensor Failure
Murderkill EstuaryDelaware Bay System (Bowers Beach, DE)
FloodDominant Estuarine
39.05 o N
75.39 o W
17 Sep 2015 to 09 Dec 2015
SeapHOx SP053, LOBO Sensors
v2.0
DIC-TA (226 samples) pHNBS (214 samples)
Murderkill EstuaryDelaware Bay System (Bowers Beach, DE)
FloodDominant Estuarine
39.05 o N
75.39 o W
11 Dec 2015 to 04 April 2016
SeapHOx SP053
v2.0
DIC-TA (116 samples) pHNBS (108 samples)
Murderkill EstuaryDelaware Bay System (Bowers Beach, DE)
FloodDominant Estuarine
39.05 o N
75.39 o W
09 May 2016 to 24 Aug 2016
SeapHOx SP053, LOBO Sensors
v3.0
DIC-TA (125 samples) pHNBS (138 samples)
Table 2. Independent reference pH, salinity constraints, and alphanumeric designations corresponding to each sensor calibration method for each reference electrode on each sampling day. Sampling Day
Reference Electrode
Independent Reference pH
Salinity Constraints
Alphanumeric Designation
01 June 2016
Internal
disc pHDIC−TA
None
1
01 June 2016
External
None
1A
01 June 2016
External
Only S>20
1B
01 June 2016
Internal
None
2
01 June 2016
External
None
2A
01 June 2016
External
Only S>20
2B
02 August 2016
Internal
None
1
02 August 2016
External
None
1
02 August 2016
Internal
None
2
02 August 2016
External
disc pHDIC−TA disc pHDIC−TA disc pHelec disc pHelec disc pHelec disc pHDIC−TA disc pHDIC−TA disc pHelec disc pHelec
None
2
80
∗ Table 3. Average values of final calibration constants (Esensor,f (T = 25℃)) applied to
all raw sensor time series. Sampling Day 01 June 2016 01 June 2016 01 June 2016 01 June 2016 01 June 2016 01 June 2016 02 August 2016 02 August 2016 02 August 2016 02 August 2016
Reference Electrode Internal External External Internal External External Internal External Internal External
Calibration Method 1 1A 1B 2 2A 2B 1 1 2 2
81
Salinity Constraints None None Only S>20 None None Only S>20 None None None None
Final Calibration Constant -0.4369 ± 0.0017 -1.4080 ± 0.0013 -1.4075 ± 0.0010 -0.4353 ± 0.0022 -1.4064 ± 0.0007 -1.4064 ± 0.0008 -0.4400 ± 0.0011 -1.4109 ± 0.0008 -0.4374 ± 0.0012 -1.4083 ± 0.0007
Table 4. Root-mean squared error (RMSE), sensor offset (c0), and sensor gain (c1) calculated from Model II least squares fits for all final sensor pH and reference pH comparisons from the 01 June 2016 sampling day. Sampling Day 01 June 2016 01 June 2016 01 June 2016 01 June 2016 01 June 2016 01 June 2016
Reference Electrode Internal External External Internal External External
Calibration Method 1 1A 1B 2 2A 2B
Salinity Constraints None None Only S>20 None None Only S>20
RMSE
c0 (intercept)
c1 (slope)
0.0275 0.0174 0.0158 0.0358 0.0114 0.0123
-0.1795±0.1683 -0.5848±0.1741 0.0882±0.3860 -0.0784±0.1592 -0.0308±0.1063 -0.1557±0.3258
1.0225±0.0210 1.0724±0.0215 0.9891±0.0473 1.0101±0.0202 1.0039±0.0132 1.0191±0.0399
82
Table 5. Root-mean squared error (RMSE), sensor offset (c0), and sensor gain (c1) calculated from Model II least squares fits for all final sensor pH and reference pH comparisons from the 02 August 2016 sampling day. Sampling Day 02 August 2016 02 August 2016 02 August 2016 02 August 2016
Reference Electrode Internal External Internal External
Calibration Method 1 1 2 2
Salinity Constraints None None None None
RMSE
c0 (intercept)
c1 (slope)
0.0159 0.0121 0.0149 0.0105
0.2281±0.1072 -0.0892±0.0814 0.3655±0.0982 0.0560±0.0689
0.9698±0.0142 1.0119±0.0108 0.9514±0.0131 0.9927±0.0092
83
Table 6. Root-mean squared error (RMSE), sensor offset (c0), and sensor gain (c1) calculated from Model II least squares ′
′
fits for pH sensor vs. pH disc property-property comparisons from the 01 June 2016 sampling day. Sampling Day 01 June 2016 01 June 2016 01 June 2016 01 June 2016 01 June 2016 01 June 2016
Reference Electrode Internal External External Internal External External
Calibration Method 1 1A 1B 2 2A 2B
Salinity Constraints None None Only S>20 None None Only S>20
RMSE
c0 (intercept)
c1 (slope)
0.0275 0.0174 0.0158 0.0358 0.0114 0.0123
-0.0160±0.0171 -0.0405±0.0129 0.0006±0.0076 -0.0068±0.0175 -0.0018±0.0085 -0.0025±0.0071
1.0225±0.0210 1.0724±0.0215 0.9891±0.0473 1.0101±0.0202 1.0039±0.0132 1.0191±0.0399
84
Table 7. Root-mean squared error (RMSE), sensor offset (c0), and sensor gain (c1) calculated from Model II least squares ′
′
fits for pH sensor vs. pH disc property-property comparisons from the 02 August 2016 sampling day. Sampling Day 02 August 2016 02 August 2016 02 August 2016 02 August 2016
Reference Electrode Internal External Internal External
Calibration Method 1 1 2 2
Salinity Constraints None None None None
RMSE
c0 (intercept)
c1 (slope)
0.0159 0.0121 0.0149 0.0105
0.0148±0.0076 -0.0051±0.0058 0.0244±0.0072 0.0045±0.0050
0.9698±0.0142 1.0019±0.0108 0.9514±0.0131 0.9927±0.0092
85
FIGURES
86 Figure 1. Different SeapHOx configurations used over the course of the project: Left Panel – v1.0, Middle Panel – v2.0, and Right Panel – v3.0.
Figure 2. Example (Top Left Panel) and extent (Top Right Panel) of sediment accumulation inside SeapHOx flow housing experienced over 2 weeks during the 02 April 2016 to 26 August 2016 SeapHOx deployment. Unsuccessful preventive measures of outfitting flow housing with multiple outflow points (Bottom Left Panel) and employing a hydrodynamic filter stuffed with fiberglass mesh (Bottom Right Panel) to combat sediment accumulation inside SeapHOx flow housing.
87
INT EXT Figure 3. Measured (A) pHraw (solid black line) and pHraw (solid blue line) during
periods of sediment accumulation inside SeapHOx flow housing and (B) turbidity (solid gold line) from 21 April 2015 to 05 May 2015.
88
Figure 4. Overview of SeapHOx field deployment method for September 2015 to August 2016 SeapHOx deployments.
89
Figure 5. Examples of biofouling observed within 1-2 weeks experienced during SeapHOx deployments from summer 2016 (Left Panel) and summer 2015 (Right Panel).
90
Figure 6. Functional implementation of the ISFET operating principle (from Martz et al., 2010).
91
disc INT EXT Figure 7. pHraw (solid blue line), pHraw (solid black line), calculated pHDIC−TA disc (squares), and calculated pHelec (diamonds) on the total hydrogen ion concentration
scale (pHT) as a function of salinity from 0900-1900 on the 01 June 2016 sampling day.
92
disc INT EXT Figure 8. pHraw (solid blue line), pHraw (solid black line), calculated pHDIC−TA disc (squares), and calculated pHelec (diamonds) on the total hydrogen ion concentration
scale (pHT) as a function of salinity from 0800-1930 on the 02 August 2016 sampling day.
93
disc disc INT Figure 9. Property-property plot of pHraw vs. pHDIC−TA (circles) and pHelec
(diamonds) as a function of salinity from the 01 June 2016 sampling day shown relative to a 1:1 (pH INT = pH disc ) relationship (dashed red line).
94
disc disc EXT Figure 10. Property-property plot of pHraw vs. pHDIC−TA (squares) and pHelec
(triangles) as a function of salinity from the 01 June 2016 sampling day shown relative to a 1:1 (pH EXT = pH disc ) relationship (dashed red line).
95
disc disc INT Figure 11. Property-property plot of pHraw vs. pHDIC−TA (circles) and pHelec
(diamonds) as a function of salinity from the 02 August 2016 sampling day shown relative to a 1:1 (pH INT = pH disc ) relationship (dashed red line).
96
disc disc EXT Figure 12. Property-property plot of pHraw vs. pHDIC−TA (squares) and pHelec
(triangles) as a function of salinity from the 02 August 2016 sampling day. shown relative to a 1:1 (pH EXT = pH disc ) relationship (dashed red line).
97
disc ∗ Figure 13. Calculated values of EINT,1 (T = 25℃) from pHDIC−TA (circles) and disc ∗ EINT,2 (T = 25℃) from pHelec (diamonds) for all measurements as a function of
salinity from the 01 June 2016 sampling day.
98
disc ∗ Figure 14. Calculated values of EEXT,1 (T = 25℃) from pHDIC−TA (squares) and disc ∗ EEXT,2 (T = 25℃) from pHelec (triangles) for all measurements as a function of
salinity from the 01 June 2016 sampling day.
99
disc ∗ Figure 15. Calculated values of EINT,1 (T = 25℃) from pHDIC−TA (circles) and disc ∗ EINT,2 (T = 25℃) from pHelec (diamonds) for all measurements as a function of
salinity from the 02 Aug 2016 sampling day.
100
disc ∗ Figure 16. Calculated values of EEXT,1 (T = 25℃) from pHDIC−TA (squares) and disc ∗ EEXT,2 (T = 25℃) from pHelec (triangles) for all measurements as a function of
salinity from the 02 August 2016 sampling day.
101
INT disc Figure 17. Property-property plot of pHfinal,1 vs. pHDIC−TA as a function of salinity
from the 01 June 2016 sampling day. The solid black line represents a Model II least INT disc squares fit and the dashed red line represents 1:1 (pHfinal,1 = pHDIC−TA ) relationship.
102
EXT disc Figure 18. Property-property plot of pHfinal,1A vs. pHDIC−TA as a function of salinity
from the 01 June 2016 sampling day. The solid black line represents a Model II least EXT disc squares fit and the dashed red line represents 1:1 (pHfinal,1A = pHDIC−TA )
relationship.
103
EXT disc Figure 19. Property-property plot of pHfinal,1B vs. pHDIC−TA as a function of salinity
from the 01 June 2016 sampling day. The solid black line represents a Model II least EXT disc squares fit and the dashed red line represents 1:1 (pHfinal,1B = pHDIC−TA )
relationship.
104
INT disc Figure 20. Property-property plot of pHfinal,2 vs. pHelec as a function of salinity from
the 01 June 2016 sampling day. The solid black line represents a Model II least INT disc squares fit and the dashed red line represents 1:1 (pHfinal,2 = pHelec ) relationship.
105
EXT disc Figure 21. Property-property plot of pHfinal,2A vs. pHelec as a function of salinity from
the 01 June 2016 sampling day. The solid black line represents a Model II least EXT disc squares fit and the dashed red line represents 1:1 (pHfinal,2A = pHelec ) relationship.
106
EXT disc Figure 22. Property-property plot of pHfinal,2B vs. pHelec as a function of salinity from
the 01 June 2016 sampling day. The solid black line represents a Model II least EXT disc squares fit and the dashed red line represents 1:1 (pHfinal,2B = pHelec ) relationship.
107
INT disc Figure 23. Property-property plot of pHfinal,1 vs. pHDIC−TA as a function of salinity
from the 02 August 2016 sampling day. The solid black line represents a Model II INT disc least squares fit and the dashed red line represents 1:1 (pHfinal,1 = pHDIC−TA )
relationship.
108
EXT disc Figure 24. Property-property plot of pHfinal,1 vs. pHDIC−TA as a function of salinity
from the 02 August 2016 sampling day. The solid black line represents a Model II EXT disc least squares fit and the dashed red line represents 1:1 (pHfinal,1 = pHDIC−TA )
relationship.
109
INT disc Figure 25. Property-property plot of pHfinal,2 vs. pHelec as a function of salinity from
the 02 August 2016 sampling day. The solid black line represents a Model II least INT disc squares fit and the dashed red line represents 1:1 (pHfinal,2 = pHelec ) relationship.
110
EXT disc Figure 26. Property-property plot of pHfinal,2 vs. pHelec as a function of salinity from
the 02 August 2016 sampling day. The solid black line represents a Model II least EXT disc squares fit and the dashed red line represents 1:1 (pHfinal,2 = pHelec ) relationship.
111
Figure 27. Murderkill Estuary-Delaware Bay System pH time-series from 09 May 2016 to 09 June 2016. (A) Raw sensor voltages for the internal (black) and external (blue) reference electrodes show four shorter-term deployments combined into one longer-term deployment. (B) pH calculated using the internal (black) and external (blue) reference electrodes.
112
Figure 28. Murderkill Estuary-Delaware Bay System pH time-series from 20 July 2016 to 24 August 2016. (A) Raw sensor voltages for the internal (black) and external (blue) reference electrodes show four shorter-term deployments combined into one longer-term deployment. (B) pH calculated using the internal (black) and external (blue) reference electrodes.
113
Figure 29. Calculated ∆pH INT−EXT anomaly (solid red line) shown relative to a zero ∆pH INT−EXT anomaly (dashed black line) and salinity (dotted green line) from 09 May 2016 to 09 June 2016. Uncharacteristically large positive ∆pH INT−EXT anomalies coincide with the first measurements of conditioning periods that follow sensor maintenance.
114
Figure 30. Calculated ∆pH INT−EXT anomaly (solid red line) shown relative to a zero ∆pH INT−EXT anomaly (dashed black line) and salinity (dotted green line) from 20 July 2016 to 24 August 2016.
115
Figure 31. Contributions of excess alkalinity to total measured alkalinity on the 01 June 2016 sampling day. (A) Excess alkalinity (ΔTA) (blue) and alkalinity calculated field from pHNBS and DIC (black) shown relative to salinity (green). (B) ΔpH due to disc contributions of excess alkalinity (blue) and pHelec assumed to be free of the
contributions of excess alkalinity (black) shown relative to salinity (green).
116
Figure 32. Contributions of excess alkalinity to total measured alkalinity on the 02 August 2016 sampling day. (A) Excess alkalinity (ΔTA) (blue) and alkalinity field calculated from pHNBS and DIC (black) shown relative to salinity (green). (B) ΔpH disc due to contributions of excess alkalinity (blue) and pHelec assumed to be free of the
contributions of excess alkalinity (black) shown relative to salinity (green).
117
Figure 33. Composite of excess alkalinity effects time-series for 09 May 2016 to 09 INT disc June 2016 SeapHOx deployment. (A) pHfinal,1 (black) calibrated using pHDIC−TA INT assumed to incorporate errors due to excess alkalinity and pHfinal,2 (green) calibrated disc INT using pHelec assumed to be free of errors due to excess alkalinity. (B) ∆pHfinal INT INT anomaly (black) calculated from difference between pHfinal,1 and pHfinal,2 shown
relative to in-situ water temperature (red).
118
Figure 34. Composite of excess alkalinity effects time-series for 20 July 2016 to 24 INT disc August 2016 SeapHOx deployment. (A) pHfinal,1 (black) calibrated using pHDIC−TA INT assumed to incorporate errors due to excess alkalinity and pHfinal,2 (green) calibrated disc INT using pHelec assumed to be free of errors due to excess alkalinity. (B) ∆pHfinal INT INT anomaly (black) calculated from difference between pHfinal,1 and pHfinal,2 shown
relative to in-situ water temperature (red).
119
Figure 35. Conditioning period at the start of the 09 May 2016 to 24 August 2016 SeapHOx deployment. Calculated ∆pHINT−EXT anomaly (solid red line) shown relative to a zero ∆pH INT−EXT anomaly (dashed black line) and salinity (dotted green line). Conditioning period starts on 9 May 2016 at 1200 (Hour 0) and goes until 12 May 2016 at 1900 (Hour 79).
120
Figure 36. Composite of intra-deployment conditioning time-series following sensor maintenance on 20 July 2016. (A) sensor measured pH INT (solid black line) and pH EXT (solid blue line) shown relative to pHT of conditioning samples calculated from Cond DIC-TA (pHDIC−TA – triangles) as a function of salinity. (B) ∆pHINT−Cond (circles)
and ∆pH EXT−Cond (squares) as a function of salinity. Conditioning period starts on 20 July 2016 at 1200 (Hour 0) and goes until 20 July 2016 at 1900 (Hour 7). Conditioning samples collected at Hour 5.5 (1730) partially froze in storage and were lost.
121
Figure 37. Composite of intra-deployment conditioning time-series following sensor maintenance on 12 August 2016. (A) sensor measured pH INT (solid black line) and pH EXT (solid blue line) shown relative to 𝑝𝐻𝑇 of conditioning samples calculated from Cond DIC-TA (pHDIC−TA – triangles) as a function of salinity. (B) ∆pHINT−Cond (circles)
and ∆pH EXT−Cond (squares) as a function of salinity. Conditioning period starts on 12 August 2016 at 1130 (Hour 0) and goes until 12 August 2016 at 1830 (Hour 7). Conditioning samples collected at Hour 5.5 (1700) were collected 7 minutes after the hour effectively missing the sensor measurement and were excluded.
122
′
′
INT disc Figure 38. Property-property plot of pHfinal,1 vs. pHDIC−TA as a function of salinity
from the 01 June 2016 sampling day. The solid black line represents a Model II least ′
′
INT disc squares fit and the dashed red line represents 1:1 (pHfinal,1 = pHDIC−TA )
relationship.
123
′
′
EXT disc Figure 39. Property-property plot of pHfinal,2 vs. pHelec as a function of salinity
from the 02 August 2016 sampling day. The solid black line represents a Model II EXT least squares fit and the dashed red line represents 1:1 (pHfinal,2
relationship.
124
′
′
disc = pHelec )
REFERENCES ACT. (2012). Alliance for Coastal Technologies: Protocol for the Verification of In Situ pH Sensors. Alin, S. R., Feely, R. A., Dickson, A. G., Hernández-Ayón, J. M., Juranek, L. W., Ohman, M. D., and Goericke, R. (2012). Robust empirical relationships for estimating the carbonate system of the southern California Current System and application to CalCOFI hydrographic cruise data (2005-2011). Journal of Geophysical Research, 117(C5). Andersson, A. J., Kline, D. I., Edmunds, P. J., Archer, S. D., Bednarsek, N., Carpenter, R. C., Chadsey, M., Goldstein, P., Grottoli, A. G., Hurst, T. P., King, A. L., Kübler, J. E., Kuffner, I. B., Mackey, K. R. M., Menge, B. A., Paytan, A., Riebesell, U., Schnetzer, A., Warner, M. E., and Zimmerman, R. C. (2015). Understanding ocean acidification impacts on organismal to ecological scales. Oceanography, 28(2). Bagshaw, E. A., Beaton, A., Wadham, J. L., Mowlem, M., Hawkings, J. R., and Tranter, M. (2016). Chemical sensors for in situ data collection in the cryosphere. Trends in Analytical Chemistry, 82, 348-357. Bates, N. R. (2007). Interannual variability of the oceanic CO2 sink in the subtropical gyre of the North Atlantic Ocean over the last 2 decades. Journal of Geophysical Research: Oceans, 112(C9). Bates, R. G. (1973). Determination of pH: Theory and Practice (2nd ed.). WileyInterscience. Bates, R. G. (1975). pH scales for seawater In E. D. Goldberg (Ed), The nature of seawater, Dahlem Konferenzen (pp. 313-338). Bates, R. G. (1982). pH measurements in the marine environment. Pure and Applied Chemistry, 54(1), 229–232. Bates, R. G., and Calais, J. G. (1981). Thermodynamics of the dissociation of bis h+ in seawater from 5 to 40c. Journal of Solution Chemistry, 10(4), 269–279. Bates, R. G., and Erickson, W. P. (1986). Thermodynamics of the dissociation of 2aminopyridinium ion in synthetic seawater and a standard for pH in marine systems. Journal of Solution Chemistry, 15(11), 891-901.
125
Bates, R. G., and Guggenheim, E. A. (1960). Report on the standardization of pH and related terminology. Pure and Applied Chemistry, 1(1), 163-168. Ben-Yaakov, S., Ruth, E., and Kaplan, I. R. (1974). Calcium carbonate saturation in northeastern Pacific: in situ determinations and geochemical implications. Deep Sea Research and Oceanographic Abstracts, 21(3), 229-243. Bresnahan Jr., P. J., Martz, T. R., Takeshita, Y., Johnson, K. S., and LaShomb, M. (2014). Best practices for autonomous measurements of seawater pH with the Honeywell Durafet. Methods in Oceanography, 9, 44-60. Bresnahan Jr., P. J., Wirth, T. R., Martz, T. R., Anderssen, A. J., Cyronak, T., D’Angelo, S., Pennise, J., Melville, W. K., Lenain, L., and Statom, N. (2016). A sensor package for mapping pH and oxygen from mobile platforms. Methods in Oceanography, 17, 1-13. Byrne, R., Kump, L., and Cantrell, K. (1988). The influence of temperature and pH on trace metal speciation in seawater. Marine Chemistry, 25(2), 163-181. Byrne, R. H. (1987). Standardization of standard buffers by visible spectrometry. Analytical Chemistry, 59(10), 1479-1481. Byrne, R. H., and Breland, J. A. (1989). High precision multiwavelength pH determinations in seawater using cresol red. Deep Sea Research Part A: Oceanographic Research Papers, 36(5), 803-810. Byrne, R. H., Mecking, S., Feely, R. A., and Liu, X. (2010). Direct observations of basin-wide acidification of the North Pacific Ocean. Geophysical Research Letters, 37(2). Cai, W. J. (2011). Estuarine and coastal ocean carbon paradox: CO2 sinks or sites of terrestrial carbon incineration? Annual Review of Marine Science, 3, 123-145. Cai, W. J., Hu., X., Huang, W. J., Jiang, L. Q., Wang, Y., Peng, T. H., and Zhang, X. (2010). Alkalinity Distributions in the western North Atlantic Ocean margins. Journal of Geophysical Research: Oceans, 115, 1-15. Cai, W. J., and Wang, Y. (1998). The chemistry, fluxes, and sources of carbon dioxide in the estuarine waters of the Satilla and Altamaha Rivers, Georgia. Limnology and Oceanography, 43(4), 657-668.
126
Cai, W. J., Wang, Y., and Hodson, R. E. (1998). Acid-base properties of dissolved organic matter in the estuarine waters of Georgia, USA. Geochimica et Cosmochimica Acta, 62(3), 473-483. Caldeira, K., and Wickett, M. (2003). Anthropogenic carbon and ocean pH. Nature, 425(6956), 365. Carter, B. R., Radich, J. A., Doyle, H. L., and Dickson, A. G. (2013). An automated system for spectrophotometric seawater pH measurements. Limnology & Oceanography: Methods, 11, 16-27. Clarke, J. S., Achterberg, E. P., Rérolle, V. M. C., Bey, S. A. K., Floquet, C. F. A., and Mowlem, M. C. (2015). Characterisation and deployment of an immobilised pH sensor spot towards surface ocean pH measurements. Analytical Chimica Acta, 897, 69-80. Clayton, T. D., and Byrne, R. H. (1993). Spectrophotometric seawater pH measurements: Total hydrogen ion concentration scale calibration of m-cresol purple and at-sea results. Deep Sea Research Part I: Oceanographic Research Papers, 40(10), 2115-2129. Clayton, T. D., Byrne, R. H., Breland, J. A., and Feely, R. A. (1995). The role of pH measurements in modern oceanic CO2-system characterization: Precision and thermodynamic consistency. Deep Sea Research Part II: Topical Studies in Oceanography, 42(2-3), 411-429. Cloern, J. E. (1996). Phytoplankton bloom dynamics in coastal ecosystems: a review with some general lessons from sustained investigation of San Francisco Bay, California. Reviews of Geophysics, 34(2), 127-168. Czerminski, J. B., Dickson, A. G., and Bates, R. G. (1982). Thermodynamics of the dissociation of morpholinium ion in seawater from 5 to 40oC. Journal of Solution Chemistry, 11(2), 79-89. DeGrandpre, M. D., Baehr, M. M., and Hammar, T. R. (1999). Calibration-free optical chemical sensors. Analytical Chemistry, 71(6), 1152-1159. DeGrandpre, M. D., Spaulding, R. S., Newton, J. O., Jaqueth, E. J., Hamblock, S. E., Umansky, A. A., and Harris, K. E. (2014). Considerations for the measurement of spectrophotometric pH for ocean acidification and other studies. Limnology and Oceanography: Methods, 12(12), 830-839.
127
de Vargas Sansalvador, I. M. P., Fay, C. D., Cleary, J., Nightingale, A. M., Mowlem, M. C., and Diamond, D. (2016). Autonomous reagent-based microfluidic pH sensor platform. Sensors and Actuators B: Chemical, 225, 369-376. deWitt, P., and Daiber, F. C. (1974). The hydrography of the Murderkill Estuary, Delaware. Chesapeake Science, 15, 84-95. DelValls, T. A., and Dickson, A. G. (1998). The pH of buffers based on 2-amino-2hydroxymethyl-1,3-propanediol (‘tris’) in synthetic sea water. Deep Sea Research Part I: Oceanographic Research Papers, 45(9), 1541-1554. Dickson, A. G. (1990). Standard potential of the reaction: AgCl (s) + 1/2 H2 (g) = Ag (s) + HCl (aq), and the standard acidity constant of the ion HSO4- in synthetic sea water from 273.15 to 318.15 K. The Journal of Chemical Thermodynamics, 22(2), 113-127. Dickson, A. G. (1993a). pH buffers for sea water media based on the total hydrogen ion concentration scale. Deep-Sea Research, 40(1), 107-118. Dickson, A. G. (1993b). The measurement of sea water pH. Marine Chemistry, 44(24), 131-142. Dickson, A. G., Camoes, M. F., Spitzer, P., Fisicaro, P., Stoica, D., Pawlowicz, R., and Feistel, R. (2016). Metrological challenges for measurements of key climatological observables. Part 3: seawater pH. Metrologia, 53(1), R26. Dickson, A. G., and Millero, F. J. (1987). A comparison of the equilibrium constants for the dissociation of carbonic acid in seawater media. Deep Sea Research Part A: Oceanographic Research Papers, 34(10), 1733-1743. Dickson, A. G., Sabine, C. L., and Christian, J. R. (2007). Guide to best practices for ocean CO2 measurements. DNREC. (2006). TMDLs for Bacteria for the Murderkill River Watershed, Delaware. Doney, S., Fabry, V., Feely, R., and Kleypas, J. (2009a). Ocean acidification: the other CO2 problem. Annual Review of Marine Science, 1, 169-192. Doney, S. C., Balch, W. M., Fabry, V. J., and Feely, R. A. (2009b). Ocean acidification: a critical emerging problem for the ocean sciences. Oceanography, 22(4), 16-25.
128
Dore, J. E., Lukas, R., Sadler, D. W., Church, M. J., and Karl, D. M. (2009). Physical and biogeochemical modulation of ocean acidification in the central North Pacific. Proceedings of the National Academy of Sciences, 106(30), 1223512240. Duarte, C. M., Hendriks, I. E., Moore, T. S., Olsen, Y. S., Steckbauer, A., Ramajo, L., Carstensen, J., Trotter, J. A., and McCulloch, M. (2013). Is Ocean Acidifcation an Open-Ocean Syndrome? Understanding Anthropogenic Impacts on Seawater pH. Estuaries and Coasts, 36, 221-236. Dzwonkowski, B., Wong, K. C., and Ullman, W. J. (2013). Water level and velocity characteristics of a salt marsh channel in the Murderkill Estuary, Delaware. Journal of Coastal Research, 30(1), 63-74. Easley, R. A., and Byrne, and R. H. (2012). Spectrophotometric calibration of pH electrodes in seawater using purified m-cresol purple. Environ. Sci. Technol., 46(9), 5018-5024. French, C. R., Carr, J. J., Dougherty, E. M., Eidson, L. A. K., Reynolds, J. C., and DeGrandpre, M. D. (2002). Spectrophotometric pH measurements of freshwater. Analytica Chimica Acta, 453, 13-20. Frieder, C. A., Nam, S. H., Martz, T. R., and Levin, L. A. (2012). High temporal and spatial variability of dissolved oxygen and pH in a nearshore California kelp forest. Biogeosciences, 9(10), 3917-3930. Friis, K., Kortzinger, A., and Wallce, D. W. R. (2004). Spectrophotometric pH measurement in the ocean: Requirements, design, and testing of autonomous charge-coupled device detector system. Limnol. Oceanogr. Methods, 2, 126136. Gabriel, M. D., Forja, J. M., Rubio, J. A., and Gómez-Parra, A. (2005). Temperature and salinity dependence of molar absorptivities of thymol blue: Application to the spectrophotometric determination of pH in estuarine waters. Ciencias Marinas, 31(1B), 309-318. González-Davila, M., Santana-Casiano, J. M., and González-Davila, E. F. (2007). Interannual variability of the upper ocean carbon cycle in the northeast Atlantic Ocean. Geophysical Research Letters, 34(7). Gran, G. (1950). Determination of the equivalence point in potentiometric titrations. Acta Chemica Scandinavica, 4, 559-577.
129
Gran, G. (1952). Determination of the equivalence point in potentiometric titrations– Part II. The Analyst, 77(920), 661-671. Hammer, K., Schneider, B., Kulinski, K., and Schulz-Bull, D. E. (2014). Precision and accuracy of spectrophotometric pH measurements at environmental conditions in the Baltic Sea. Estuarine, Coastal, and Shelf Science, 146, 24-32. Hansson, I. (1973a). A new set of acidity constants for carbonic acid and boric acid in sea water. Deep Sea Research and Oceanographic Abstracts, 20(5), 461-478. Hansson, I. (1973b). A new set of pH-scales and standard buffers for sea water. Deep Sea Research and Oceanographic Abstracts, 20(5), 479-491. Harris, D. C. (2007). Quantitative Chemical Analysis (7th ed.). W.H. Freeman and Company. Hawley, J. E., and Pytkowicz, R. M. (1973). Interpretation of pH measurements in concentrated electrolyte solutions. Marine Chemistry, 1(3), 245-250. Head, P. C. (1985). Practical estuarine chemistry: a handbook. Cambridge University Press. Hofmann, G. E., Smith, J. E., Johnson, K. S., Send, U., Levin, L. A., Micheli, F., Paytan, A., Price, N. N., Peterson, B., Takeshita, Y., Matson, P. G., DerseCrook, E., Kroeker, K. J., Gambi, M., Rivest, E. B., Frieder, C. A., Yu, P. C., and Martz, T. R. (2011). High-frequency dynamics of ocean pH: a mutliecosystem comparison. PLoS One, 6. Hopkins, A. E., Sell, K. S., Soli, A. L., and Byrne, R. H. (2000). In-situ spectrophotometric pH measurements: the effects of pressure on thymol blue protonation and absorbance characteristics. Marine Chemistry, 71(1-2), 103109. Huang, W. J., Wang, Y., and Cai, W. J. (2012). Assessment of sample storage techniques for total alkalinity and dissolved inorganic carbon in seawater. Limnology and Oceanography: Methods, 10(9), 711-717. Hulanicki, A., Glab, S., and Ingman, F. (1991). Chemical Sensors: Definitions and Classification. Journal of Pure & Applied Chemistry, 63(9), 1247-1250. Janata, J. (2009). Chemical Sensors (2nd ed.). Springer-Verlag Press.
130
Jiang, L. Q., Cai, W. J., Wannikhoff, R., Wang, Y., and Luger, H. (2008). Air-sea CO2 fluxes on the US South Atlantic Bight: spatial and seasonal variability. Journal of Geophysical Research: Oceans (1978-2012), 113(C7). Joesoef, A., Huang, W. J., Gao, Y., and Cai, W. J. (2015). Air-water fluxes and sources of carbon dioxide in the Delaware Estuary: spatial and seasonal variability. Biogeosciences, 12, 6085-6101. Johnson, K. S., Jannasch, H. W., Coletti, L. J., Elrod, V. A., Martz, T. R., Takeshita, Y., Carlson, R. J., and Connery, J. G. (2016). Deep-Sea DuraFET: A pressure tolerant pH sensor designed for global sensor networks. Analytical Chemistry, 88(6), 3249-3256. Johnson, K. S., Voll, R., Curtis, C. A., and Pytkowicz, R. M. (1977). A critical examination of the NBS pH Scale and the determination of titration alkalinity. Deep-Sea Research, 24(10), 915-926. Kent County Board of Public Works (KCBPW). (2012). Kent County Regional Wastewater Treatment Facilities. http://www.co.kent.de.us/media/804272/WTF-FACT-SHEET-2013_1.pdf (accessed 25.02.2016). Khoo, K. H., Ramette, R. W., Culberson, C. H., and Bates, R. G. (1977). Determination of hydrogen ion concentrations in seawater from 5 to 40oC: Standard potentials at salinities from 20 to 45. Analytical Chemistry, 49(1), 2934. Kim H. C., and Lee, K. (2009). Significant contribution of dissolved organic matter to seawater alkalinity. Geophysical Research Letters, 36(20). King, D., and Kester, D. R. (1989). Determination of seawater pH from 1.5 to 8.5 using coulometric indicators. Marine Chemistry, 26(1), 5-20. Kleypas, J., Feely, R., Fabry, V., Langdon, C., Sabine, C., and Robbins, L. (2006). Impacts of ocean acidification on coral reefs and other marine calcifiers: A guide for future research, report of a workshop held by NSF. NOAA, and the USGS.
131
Kline, D. I., Teneva, L., Schneider, K., Miard, T., Chai, A., Marker, M., Headley, K., Opdyke, B., Nash, M., Valetich, M., Caves, J. K., Russell, B. D., Connell, S. D., Kirkwood, B. J., Brewer, P., Peltzer, E., Silverman, J., Caldeira, K., Dunbar, R. B., Koseff, J. R., Monismith, S. G., Mitchell, B. G., Dove, S., and Hoegh-Guldberg, O. (2012). A short-term in situ CO2 enrichment experiment on Heron Island (GBR). Scientific Reports, 2, 413. Ko, Y. H., Lee, K., Eom, K. H., and Han, I. S. (2016). Organic alkalinity produced by phytoplankton and its effect on the computation of the ocean carbon parameters. Limnology and Oceanography. Lai, C. Z., DeGrandpre, M. D., Wasser, B. D., Brandon, T. A., Clucas, D. S., Jaqueth, E. J., Benson, Z. D., Beatty, C. M., and Spaulding, R. S. (2016). Spectrophotometric measurement of freshwater pH with purified meta-cresol purple and phenol red. Limnology and Oceanography: Methods. Larsen, M., Borisov, S. M., Grunwald, B., Klimant, I., and Glud, R. N. (2011). A simple and inexpensive high resolution color ratiometric planar optode imaging approach: applications to oxygen and pH sensing. Limnology and Oceanography: Methods, 9(9), 348-360. Lee, K., Kim, T. W., Byrne, R. H., Millero, F. J., Feely, R. A., and Liu, Y. M. (2010). The universal ratio of boron to chlorinity for the North Pacific and North Atlantic oceans. Geochimica et Cosmochimica Acta, 74(6), 1801-1811. Liebsch, G., Kilmant, I., Krause, C., and Wolfbeis, O. S. (2001). Fluorescent imaging of pH with optical sensors using time domain dual lifetime referencing. Analytical Chemistry, 73(17), 4354-4363. Liu, X., Patsavas, M. C., and Byrne, R. H. (2011). Purification and characterization of meta-cresol purple for spectrophotometric seawater pH measurements. Environ. Sci. Technol., 45(11), 4862-4868. Loucaides, S., Tyrrell, T., Achterberg, E. P., Torres, R., Nightingale, P. D., Kitidis, V., Serret, P., Woodward, M., and Robinson, C. (2012). Biological and physical forcing of carbonate chemistry in an upwelling filament off northwest Africa: Results from a Lagrangian study. Global Biogeochemical Cycles, 26(3). MacLeod, C. D., Doyle, H. L., and Currie, K. I. (2015). Technical Note: Maximising accuracy and minimizing cost of a potentially regulated ocean acidification simulation system. Biogeosciences, 12(3), 713-721.
132
Marion, G. M., Millero, F. J., Camoes, M. F., Spitzer, P., Feistel, R., and Chen, C. T. (2011). pH of seawater. Marine Chemistry, 126(1-4), 89-96. Martz, T. (2012). SeaFET & SeapHOx Users Guide. University of California, San Diego. Scripps Institute of Oceanography. Martz, T., Send, U., Ohman, M. D., Takeshita, Y., Bresnahan Jr., P. J., Kim, H. J., and Nam, S. (2014). Dynamic variability of biogeochemical ratios in the Southern California Current System. Geophysical Research Letters, 41(7), 2496-2501. Martz, T. R., Carr, J. J., French, C. R., and DeGrandpre, M. D. (2003). A submersible autonomous sensor for spectrophotometric pH measurements of natural waters. Analytica Chimica Acta, 75, 1844-1850. Martz, T. R., Connery, J. G., and Johnson, K. S. (2010). Testing the Honeywell Durafet for seawater pH applications. Limnology & Oceangraphy: Methods, 8, 172-184. Matson, P. G., Martz, T. R., and Hofmann, G. E. (2011). High-frequency observations under Antarctic sea ice in the southern Ross Sea. Antarctic Science, 23(06), 607-613. McClendon, J. F. (1917). The standardization of a new colorimetric method for the determination of the hydrogen ion concentration, CO2 tension, and CO2 and O2 content of seawater, of animal heat, and of CO2 of the air, with a summary of similar data on bicarbonate solutions in general. Journal of Biological Chemistry, 30(2), 265-288. Mehrbach, C., Culberson, C. H., Hawley, J. E., and Pytkowicz, R. M. (1973). Measurement of the apparent dissociation constants of carbonic acid in seawater at atmospheric pressure. Limnology and Oceanography, 18(6), 897907. Midorikawa, T., Ishii, M., Saito, S., Sasano, D., Kosugi, N., Motoi, T., Kamiya, H., Nakadate, A., Nemoto, K., and Inoue, H. Y. (2010). Decreasing pH trend estimated from 25-yr time series of carbonate parameters in the western North Pacific. Tellus B, 62(5), 649-659. Mikhelson, K. N. (2013). Ion-Selective Electrodes. Springer-Verlag Press. Millero, F. J. (1986). The pH of estuarine waters. Limnology and Oceanography, 31(4), 839–847.
133
Millero, F. J., Graham, T. B., Huang, F., Bustos-Serrano, H., and Pierrot, D. (2006). Dissociation constants of carbonic acid in seawater as a function of salinity and temperature. Marine Chemistry, 100(1), 80-94. Millero, F. J., Zhang, J. Z., Fiol, S., Sotolongo, S., Roy, R. N., Lee, K., and Mane, S. (1993). The use of buffers to measure pH for seawater. Marine Chemistry, 22(2-4), 143-152. Mosley, L. M., Husheer, S. L. G., and Hunter, K. A. (2004). Spectrophotometric pH measurement in estuaries using thymol blue and m-cresol purple. Marine Chemistry, 91(1-4), 175-186. Nemzer B. V., and Dickson, A. G. (2005). The stability and reproducibility of tris buffers in synthetic seawater. Marine Chemistry, 96(3-4), 237-242. Newton, J. A., Feely, R. A., Jewett, E. B., Williamson, P., and Mathis, J. (2015). Global Ocean Acidification Observing Network: Requirements and Governance Plan. 2nd ed., GOA-ON. Nightingale, A. M., Beaton A. D., and Mowlem, M. C. (2015). Trends in microfluidic systems for in situ chemical analysis of natural waters. Sensors and Actuators B: Chemical, 221, 1398-1405. Orr, J. C., Fabry, V. J., Aumont, O., Bopp, L., Doney, S. C., Feely, R. A., Gnanadesikan, A., Gruber, N., Ishida, A., Joos, F., Key, R. M., Lindsay, K., Maier-Reimer, E., Matear, R., Monfray, P., Mouchet, A., Najjar, R. G., Plattner, G. K., Rodgers, K. B., Sabine, C. L., Sarmiento, J. L., Schlitzer, R., Slater, R. D., Totterdell, I. J., Weirig, M. F., Yamanaka, Y., and Yool, A. (2005). Anthropogenic ocean acidification over the twenty-first century and its impacts on calcifying organisms. Nature, 437(7059), 681-686. Papadimitriou, S., Loucaides, S., Réroll,e V., Achterberg, E. P., Dickson, A. G., Mowlem, M., and Kennedy, H. (2016). The measurement of pH in saline and hypersaline media at sub-zero temperatures: Characterization of Tris buffers. Marine Chemistry. Patsavas, M. C., Byrne, R. H., and Liu, X. (2013a). Purification of meta-cresol purple and cresol red by flash chromatography: Procedures for ensuring accurate spectrophotometric seawater pH measurements. Marine Chemistry, 150, 19-24. Patsavas, M. C., Byrne, R. H., and Liu, X. (2013b). Physical-chemical characterization of purified cresol red for spectrophotometric pH measurements in seawater. Marine Chemistry, 155, 158-164.
134
Patsavas, M. C., Byrne, R. H., Wanninkhof, R., Feely, R. A., and Cai, W. J. (2015). Internal consistency of marine carbonate system measurements and assessments of aragonite saturation state: Insights from two US coastal cruises. Marine Chemistry, 176, 9-20. Peltzer, E. (2007). Model II least squares fit: lsqfitma.m. http://www.mbari.org/staff/etp3/regress/lsqfitma.m. Pierrot, D., Lewis, E., and Wallace, D. W. R. (2006). CO2SYS DOS Program Developed for CO2 System Calculations. ORNL/CDIAC-105, Carbon Dioxide Information Analysis Center, Oak Ridge National Laboratory, US Department of Energy, Oak Ridge, TN. Pratt, K. W. (2014). Measurement of pHT values of TRIS Buffers in artificial seawater at varying mole rations of TRIS:TRIS-HCl. Marine Chemistry, 162, 89-95. Price, N. N., Martz, T. R., Brainard, R. E., and Smith, J. E. (2012). Diel variability in sewatet pH relates to calcification and benthic community structure on coral reefs. PloS one, 7(8), e43843. Pytkowicz, R. M., Ingle, S. E., and Mehrbach, C. (1974). Invariance of apparent of equilibrium constants with pH. Limnology and Oceanography, 19(4), 665-669. Raghuraman, B., Gustavson, G., Van Hal, R. E. G., Dressaire, E., and Zhdaneev, O. (2006). Extended-range spectroscopic pH measurement using optimized mixtures of dyes. Applied Spectroscopy, 60(12), 1461-1468. Ramette, R. W., Culberson, C. H., and Bates, R. G. (1977). Acid-base properties of tris(hydroxymethyl) aminomethance (tris) buffers in sea water from 5 to 40oC. Analytical Chemistry, 49(6), 867-870. Raven, J., Caldeira, K., Eldersfield, H., Hoegh-Guldberg, O., Liss, P., Riebesell, U., Shepherd, J., Turley, C., and Watson, A. (2005). Ocean acidification due to increasing atmospheric carbon dioxide. The Royal Society. Rérolle, V., Ruiz-Pino, D., Rafizadeh, M., Loucaides, S., Papadimitriou, S., Mowlem, M., and Chen, J. (2016). Measuring pH in the Arctic Ocean: Colorimetric method or SeaFET? Methods in Oceanography, 17, 32-49. Rérolle, V. M. C., Floquet, C. F. A., Harris, A. J., Mowlem, M. C., Bellerby, R. R., and Achterberg, E. P. (2013). Development of a colorimetric microfluidic pH sensor for autonomous seawater measurements. Analytica Chimica Acta, 786, 124-131.
135
Rérolle, V. M. C., Floquet, C. F. A., Mowlem, M. C., Bellerby, R. R., Connelly, D. P., and Achterberg, E. P. (2012). Seawater-pH measurements for oceanacidification observations. Trends in Analytical Chemistry, 40, 146-157. Ribas-Ribas, M., Rérolle, V., Bakker, D. C. E., Kitidis, V., Lee, G., Brown, I., Achterberg, E. P., Hardman-Mountford, N., and Tyrrell, T. (2014). Intercomparison of carbonate chemistry measurements on a cruise in northwestern European shelf seas. Biogeosciences, 11, 4339-4355. Rivest, E. B., O’Brien, M., Kapsenberg, L., Gotschalk, C. C., Blanchette, C. A., Hoshijima, U., and Hofmann, G. E. (2016). Beyond the benchtop and the benthos: Dataset management planning and design for time series of ocean carbonate chemistry associated with Durafet-based pH sensors. Ecological Informatics. Robert-Baldo, G. L., Morris, M. J., and Byrne, R. H. (1985). Spectrophotometric determination of seawater pH using phenol red. Analytical Chemistry, 57(13), 2564-2567. Rodriguez, C, Huang, F, and Millero, F. J. (2015) The partial molal volume and compressibility of Tris and Tris-HCl in water and 0.725 m NaCl as a function of temperature. Deep Sea Research Papers: Part I: Oceanographic Research Papers, 104, 41-51. Sandifer, J., and Voycheck, J. (1999). A review of biosensor and industrial applications of pH-ISFETs and an evaluation of Honeywell’s “DuraFET.” Microchimica Acta, 131, 91-98. Seidel, M. P., DeGrandpre, M. D., and Dickson, A. G. (2008). A sensor for in situ indicator-based measurements of seawater pH. Marine Chemistry, 109(1-2), 18-28. Sharp, J. H., Yoshiyama, K., Parker, A. E., Schwartz, M. C., Curless, S., Beauregard, A. Y., Ossolinski, J. E., and Davis, A. R. (2009). A biogeochemical view of estuarine eutrophication: lessons from the Delaware Estuary. Estuaries and Coasts, 32(6), 1023-1043. Sillen, G. (1968). Master variables and activity scales. Skoog, D., Holler, F. J., and Crouch, S. R. (2006). Principles of Instrumental Analysis (6th ed.). Cengage Learning, Inc.
136
Soli, A. L., Brody, J. P., and Byrne, R. H. (2013). The effect of pressure on metaCresol Purple protonation and absorbance characteristics for spectrophotometric pH measurements in seawater. Marine Chemistry, 157, 162-169. Sørensen, S. P. L. (1909). Enzymstudien. II. Mitteilung. Uber die Messung and die Bedeutung de Wasserstoffionenkonzentration bei enzymatische Prozessen. Biochem. Z., 21, 131-304. Sørensen S. P. L., and Linderstrøm-Lang, K. (1924). On the determination and values of π0 in electromotoric measurements of hydrogen ion concentrations. Compt. rend. du Lab. Carlsberg, 15, 1-40. Takeshita Y., Martz, T. R., Coletti, L. J., Dickson, A. G., Jannasch, H. W., and Johnson, K. S. (2016). The effects of pressure on pH of Tris buffer in synthetic seawater. Marine Chemistry. Takeshita, Y., Martz, T. R., Johnson, K. S., and Dickson, A. G. (2014). Characterization of an Ion-Selective Field Effect Transistor and Chloride IonSelective Electrodes for pH Measurements in Seawater. Analytical Chemistry, 86(22), 1189-1195. Ullman, W. J., Aufdenkampe, A., Hays, R. L., and Dix, S. (2013). Nutrient Exchange between a Salt Marsh and the Murderkill Estuary, Kent County, Delaware. Voynova, Y. G., Lebaron, K. C., Barnes, R. T., and Ullman, W. J. (2015). In situ response of bay productivity to nutrient loading from a small tributary: The Delaware Bay-Murderkill Estuary tidally-coupled biogeochemical reactor. Estuarine, Coastal, and Shelf Science, 160, 33-48. Weisberg, S. B., Bednarsek, N., Feely, R. A., Chan, F., Boehm, A. B., Sutula, M., Ruesink, J. L., Hales, B., Largier, J. L., and Newton, J. A. (2016). Water quality criteria for an acidifying ocean: Challenges and opportunities for improvement. Ocean & Coastal Management, 126, 31-41. Whitfield, M., Butler, R. A., and Covington, A. K. (1985). The determination of pH in estuarine waters I. Definition of pH scales and selection of buffers. Oceanological Acta, 8(4), 423-432. Wong, K. C., Dzwonkowski, B., and Ullman, W. J. (2009). Temporal and spatial variability of sea level and volume flux in the Murderkill Estuary. Estuarine, Coastal, and Shelf Science, 84, 440-446.
137
Yang, B., Byrne, R. H., and Lindemuth, M. (2015). Contributions of organic alkalinity to total alkalinity in coastal waters: A spectrophotometric approach. Marine Chemistry, 176, 199-207. Yao, W., and Byrne, R. H. (1998). Simplified seawater alkalinity analysis: Use of linear array spectrometers. Deep Sea Research Part I: Oceanographic Research Papers, 45(8), 1383-1392. Yao, W., and Byrne, R. H. (2001). Spectrophotometric determination of freshwater pH using bromocresol purple and phenol red. Environ. Sci. Technol., 35(6), 11971201. Yao, W., Liu, X., and Byrne, R. H. (2007). Impurities in indicators used for spectrophotometric seawater pH measurements: Assessment and remedies. Marine Chemistry, 107(2), 167-172. Zeebe, R. E., and Wolf-Gladrow, D. A. (2001). CO2 in seawater: equilibrium, kinetics, isotopes (No. 65). Gulf Professional Publishing. Zhang, H., and Byrne, R. H. (1996). Spectrophotometric pH measurements of surface seawater at in-situ conditions: Absorbance and protonation behavior of thymol blue. Marine Chemistry, 52(1), 17-25.
138
